Tomografia Fotoacustica 2015!! PDF

Aperture
Fixed Mirror
Transducer
MEMS Mirror
Plastic Film
Photoacoustic
To m o g r a p h y
HUABEI JIANG
Photoacoustic
To m o g r a p h y
HUABEI JIANG
University of Florida, Gainesville, FL, USA
Boca Raton London New York
CRC Press is an imprint of the

Taylor & Francis Group, an informa business
CRC Press
Taylor & Francis Group
6000 Broken Sound Parkway NW, Suite 300
Boca Raton, FL 33487-2742
© 2015 by Taylor & Francis Group, LLC
CRC Press is an imprint of Taylor & Francis Group, an Informa business
No claim to original U.S. Government works

Version Date: 20141103
International Standard Book Number-13: 978-1-4822-6104-2 (eBook - PDF)
This book contains information obtained from authentic and highly regarded sources. Reasonable
efforts have been made to publish reliable data and information, but the author and publisher cannot
assume responsibility for the validity of all materials or the consequences of their use. The authors and
publishers have attempted to trace the copyright holders of all material reproduced in this publication
and apologize to copyright holders if permission to publish in this form has not been obtained. If any
copyright material has not been acknowledged please write and let us know so we may rectify in any
future reprint.
Except as permitted under U.S. Copyright Law, no part of this book may be reprinted, reproduced,
transmitted, or utilized in any form by any electronic, mechanical, or other means, now known or
hereafter invented, including photocopying, microfilming, and recording, or in any information stor-
age or retrieval system, without written permission from the publishers.
For permission to photocopy or use material electronically from this work, please access www.copy-
right.com (http://www.copyright.com/) or contact the Copyright Clearance Center, Inc. (CCC), 222
Rosewood Drive, Danvers, MA 01923, 978-750-8400. CCC is a not-for-profit organization that pro-
vides licenses and registration for a variety of users. For organizations that have been granted a photo-
copy license by the CCC, a separate system of payment has been arranged.
Trademark Notice: Product or corporate names may be trademarks or registered trademarks, and are
used only for identification and explanation without intent to infringe.
Visit the Taylor & Francis Web site at
http://www.taylorandfrancis.com
and the CRC Press Web site at
http://www.crcpress.com
To Yonghong, Max, and Michael
Contents
Preface.......................................................................................................................xi
About the Author.................................................................................................... xiii
Chapter 1 Fundamentals of Photoacoustic Tomography....................................... 1

1.1 Photoacoustic Effect...................................................................1
1.1.1 Empirical Description................................................... 1
1.1.2 Rigorous Theory...........................................................4
1.2 Image Reconstruction Methods.................................................. 5
1.2.1 Delay-and-Sum Beam Forming Algorithm................... 5
1.2.2 Iterative Nonlinear Algorithm....................................... 6
1.2.2.1 Frequency-Domain FE-Based Algorithm.......... 6
1.2.2.2 Time-Domain FE-Based Algorithm.................. 8
1.2.3 A-Line/B-Mode Image Formation Method................. 11
1.3 Instrumentation........................................................................ 13
Chapter 2 Quantitative Photoacoustic Tomography............................................ 15

2.1 Recovery of Optical Absorption Coefficient: Method 1.......... 15
2.2 Recovery of Optical Absorption Coefficient: Method 2..........20
2.3 Recovery of Optical Absorption Coefficient: Method 3..........24
2.4 Recovery of Optical Absorption Coefficient: Method 4.......... 29
2.5 Simultaneous Recovery of Optical Absorption
Coefficient and Acoustic Velocity............................................34
2.6 Multispectral PAT.................................................................... 41
Chapter 3 Image Enhancement Software and Hardware Approaches................. 47

3.1 Dual Mesh Scheme................................................................... 47
3.2 Adjoint Sensitivity Method....................................................... 49
3.3 Total-Variation-Minimization Scheme..................................... 50
3.3.1 Mathematical Derivations........................................... 51
3.3.2 Examples: Reduction of Noise Effect......................... 53
3.3.2.1 Simulations.................................................. 53
3.3.2.2 Phantom Experiments.................................. 58
3.3.3 Examples: Image Reconstruction from Few-
Detector and Limited-Angle Data............................... 62
3.4 Radiative Transfer Equation..................................................... 67
3.4.1 The RTE and Its Finite Element Discretization.......... 69
3.4.2 The RTE-Based Quantitative PAT.............................. 72
3.4.3 Results and Discussion................................................ 73
v
vi Contents
3.4.3.1 Comparison of the Transport and

Diffusion Calculations................................. 73
3.4.3.2 Comparison of the DE- and RTE-
Based qPAT................................................. 75
3.5 Parallel Computation................................................................ 83
3.6 Variable-Thickness Multilayered Polyvinylidene Fluoride
Transducer................................................................................ 85
3.7 AlN-Based Piezoelectric Micromachined Ultrasonic
Transducer................................................................................ 91
3.8 Liquid Acoustic Lens...............................................................97
3.9 Microelectromechanical Systems Scanning Mirror............... 103
3.9.1 Photoacoustic Imaging System................................. 103
3.9.2 Results and Discussion.............................................. 105
Chapter 4 Transducer Array-Based Photoacoustic Tomography: 2D, 3D,

and 4-D Photoacoustic Imaging ....................................................... 109
4.1 Array-Based PAT System and 2D Imaging............................ 109
4.1.1 Laser Source.............................................................. 111
4.1.2 Full-Ring Ultrasound Transducer Array................... 112
4.1.3 Control Electronics and Data Acquisition................. 112
4.1.4 System Evaluation and Experimental Studies........... 114
4.1.4.1 System Calibration..................................... 114
4.1.4.2 Spatial Resolution...................................... 116
4.1.4.3 Phantom Evaluation................................... 117
4.2 3D Imaging............................................................................. 119
4.2.1 System Description................................................... 119
4.2.2 Phantom Experiments............................................... 121
4.2.3 In Vivo Experiments................................................. 123
4.3 4-D Imaging........................................................................... 124
4.3.1 Drug Delivery Monitoring........................................ 126
4.3.2 Tumor Therapy Monitoring....................................... 126
Chapter 5 Photoacoustic Microscopy and Photoacoustic Computed

Microscopy........................................................................................ 131
5.1 Optical-Resolution Photoacoustic Microscopy...................... 131
5.2 Acoustic-Resolution Photoacoustic Microscopy.................... 132
5.3 C-Scan Photoacoustic Microscopy......................................... 135
5.3.1 Materials and Methods.............................................. 135
5.3.1.1 Experimental Setup................................... 135
5.3.1.2 Image Reconstruction Method.................. 136
5.3.1.3 Phantom Preparation................................. 136
5.3.1.4 Animal Preparation and Histological
Sectioning.................................................. 136
Contents vii
5.3.2.1 Phantom Experiments................................ 137

5.3.2.2 In Vivo Experiments.................................. 138
5.4 Photoacoustic Computed Microscopy.................................... 140
5.4.1 Methods..................................................................... 141
5.4.1.1 PAM Imaging System................................ 141
5.4.1.2 Phantom Experiments................................ 142
5.4.1.3 Animal Experiments.................................. 142
5.4.1.4 Image Reconstruction................................ 143
5.4.2 Results....................................................................... 145
5.4.3 Discussion................................................................. 149
5.5 Photoacoustic Microscopy Based on Acoustic Lens with
Variable Focal Length............................................................ 155
5.6 Confocal Photoacoustic Microscopy Using a Single
Multifunctional Lens.............................................................. 160
Chapter 6 Multimodal Approaches.................................................................... 169

6.1 PAT/DOT................................................................................ 169
6.1.1 Material and Methods............................................... 169
6.1.1.1 System Description.................................... 169
6.1.1.2 Quantitative Reconstruction Algorithms....172
6.1.2.1 Spatial Resolution of the System............... 173
6.1.2.2 PAT Performance....................................... 173
6.1.2.3 PAT and DOT Comparison....................... 175
6.1.2.4 Ex Vivo Experiment.................................. 177
6.2 PAT/FMT................................................................................ 179
6.2.1 Methods..................................................................... 180
6.2.2.1 Comparison of Image Pattern.................... 183
6.2.2.2 Comparison of Spatial Resolution............. 184
6.2.2.3 Comparison of Sensitivity......................... 185
6.3 PAT/Ultrasound...................................................................... 186
6.4 Optical-Resolution PAM/OCT............................................... 188
6.4.1 Materials and Methods.............................................. 188
Chapter 7 Contrast Agents–Based Molecular Photoacoustic Tomography....... 193

7.1 Gold Nanoparticles................................................................. 193
7.2 Graphene Nanosheets............................................................. 196
7.3 Urokinase Plasminogen Activator Receptor (uPAR)–
Targeted Magnetic Iron Oxide Nanoparticles (NIR830-
ATF-IONP)............................................................................. 198
7.3.1 Material and Methods............................................... 198
7.3.1.1 Cell Line.................................................... 198
viii Contents
7.3.1.2 Preparation of NIR830-ATF-IONP

and Control NIR830-Bovine Serum
Albumin-IONP (NIR830-BSA-IONP)...... 198
7.3.1.3 Animal Tumor Model................................200
7.3.1.4 Photoacoustic Microscopy Imaging
System........................................................200
7.3.1.5 Near-Infrared Planar Fluorescence
Imaging System......................................... 201
7.3.1.6 Image Processing....................................... 201
7.3.1.7 Histological Analysis................................. 201
7.3.1.8 Statistical Analysis....................................202
7.3.2 Results.......................................................................202
7.3.3 Discussion.................................................................205
7.4 HER-2/neu Targeted Magnetic Iron Oxide Nanoparticles
for Dual-Modal Photoacoustic and Fluorescence
Molecular Tomography (PAT/FMT) of Ovarian Cancer.......207
7.4.1 Methods.....................................................................208
7.4.1.1 Cell Lines...................................................208
7.4.1.2 HER-2/neu Specific Affibody
Conjugation to IONP.................................208
7.4.1.3 Preparation of NIR-Bovine Serum
Albumin-IONP (NIR-830-BSA-IONPs)... 210
7.4.1.4 Evaluation of Labeling Efficiency by
Prussian Blue Staining............................... 210
7.4.1.5 Orthotopic Human Ovarian Cancer
Xenograft Model........................................ 210
7.4.1.6 Fluorescence Molecular Tomography
Imaging System......................................... 211
7.4.1.7 Photoacoustic Imaging System.................. 211
7.4.1.8 In Vivo Planar Fluorescence Imaging....... 212
7.4.1.9 In Vivo Imaging in Animal Tumor
Models....................................................... 212
7.4.1.10 Histology Analysis..................................... 212
7.4.1.11 Statistical Analysis.................................... 213
7.4.2 Results....................................................................... 213
7.4.3 Discussion................................................................. 219
Chapter 8 Clinical Applications and Animal Studies........................................ 223

8.1 Joint Imaging.......................................................................... 223
8.1.1 2D Single-Spectral Quantitative PAT.......................224
8.1.1.1 Materials and Methods..............................224
8.1.1.2 Results and Discussion.............................. 226
8.1.2 2D Multispectral Quantitative PAT........................... 228
8.1.2.1 Materials and Methods.............................. 228
Contents ix
8.1.3 3D Single-Spectral Quantitative PAT....................... 233

8.1.4 3D Multispectral Quantitative PAT........................... 237
8.2 Intraoperative Imaging........................................................... 242
8.2.1 Intraoperative Photoacoutc Tomography (iPAT)
System....................................................................... 243
8.2.2 Results and Discussion..............................................244
8.3 Brain Imaging.........................................................................248
8.3.1 Methods..................................................................... 250
8.3.1.1 Animals..................................................... 250
8.3.1.2 PAT Imaging.............................................. 250
8.3.1.3 Electrode Implantation Surgery................. 250
8.3.1.4 Induction of Seizures................................. 251
8.3.1.5 Image Analysis.......................................... 251
8.3.2.1 Noninvasive Epileptic Foci Localization....251
8.3.2.2 Real-Time Monitoring of Epileptic
Events......................................................... 252
8.3.2.3 Dynamical Changes of Vasculature
during Interictal Discharges...................... 256
8.4 Imaging of Tumor Vasculature Development......................... 258
8.4.1 Methods..................................................................... 258
8.4.1.1 Photoacoustic Imaging System.................. 258
8.4.1.2 Mouse Breast Cancer Xenograft Models..... 258
8.5 Intravascular Imaging............................................................. 262
8.6 Breast Imaging.......................................................................264
8.6.1 Photoacoustic Tomography System...........................266
8.6.2 Clinical Experiments................................................. 267
8.6.2.1 PAT for Breast Cancer Detection.............. 267
8.6.2.2 PAT for Neoadjuvant Monitoring.............. 269
Bibliography.......................................................................................................... 273
Preface
Photoacoustic tomography (PAT) is an emerging biomedical imaging modality that
combines the high-contrast and spectroscopic-based specificity of optical imaging
with the high spatial resolution of ultrasound imaging in a single modality. By lis-
tening to light, PAT detects tissue-absorbed photons ultrasonically through the pho-
toacoustic effect. Since ultrasonic scattering is two to three orders of magnitude
weaker than optical scattering in tissue, PAT breaks the 2–4 mm spatial resolution
limit associated with pure optical tomography such as diffuse optical tomography
(DOT) for deep tissue imaging, or the ~1 mm depth limit associated with confo-
cal and multiphoton microscopy and optical coherent tomography (OCT). In PAT,
tissue is excited with a short (typically a few nanoseconds) laser pulse (focused or
unfocused); the subsequent laser-induced transient photoacoustic waves in the range
of 1–100 MHz, due to the transient thermoelastic expansion of light-absorbing
components in tissue, are detected by wideband unfocused or focused ultrasound
transducer(s). Unique advantages of PAT are that functional or biochemical parame-
ters such as deoxy-hemoglobin (HbR), oxy-hemoglobin (HbO2), water (H2O), lipids,
and so forth along with vasculature and blood flow can be imaged in high resolution.
In addition, highly specific molecular PAT can be realized through the use of molec-
ular contrast agents. Finally, PAT can be made portable for bedside applications, is
economical, and uses non-ionization radiation.
PAT has found its potential clinical applications in several areas. In breast imag-
ing, PAT offers the submillimeter-resolution ability to quantitatively image the high
optical contrast generated through the presence of blood, water, and lipid, which are
the predominant transformations associated with malignancy. Clinical studies con-
ducted at multiple institutions and countries have repeatedly shown that 2-to-11 and
higher absorption contrasts exist in breast cancers that can be imaged by PAT. These
studies suggest that PAT has the potential to detect breast tumors at early stages.
Application of PAT to joint imaging has been recently explored, offering an oppor-
tunity for early detection and monitoring of progressive diseases including osteoar-
thritis and rheumatoid arthritis. In this case, the optical contrast is produced through
the degraded articular cartilage and the increased water content and turbidity in
the synovial cavity. Other clinical applications of PAT also are starting to appear,
including intravascular imaging and intraoperative imaging during cancer treatment.
While brain imaging and electrophysiology play a central role in neuroscience
research and in the evaluation of neurological disorders, a single noninvasive modal-
ity that offers both high spatial and temporal resolution is currently not available.
PAT may become such a neuroimaging modality at least in animals. In Chapter 8,
I show an acute epilepsy rat model in which PAT can noninvasively track seizure
brain dynamics with both high spatial and temporal resolution and at a depth that is
clinically relevant. The noninvasive yet whole surface and depth capabilities of PAT
actually seeing what is happening during ictogenesis in terms of seizure onset and
spread.
xi
xii Preface
The concept of PAT emerged in the mid-1990s, and the field of PAT is now rap-
idly moving forward. Several reviews and edited books on PAT have been published.
However, I have decided to write a book that presents a detailed and comprehensive
treatment of PAT. It appears timely to produce such a book for the first time in this
field. The book is essentially a collection of the research work that the author and
his colleagues have been pursuing over the past decade. This actually presents a
unique feature of the book as it allows the principles of PAT and its applications
to be treated in a systematic way. In addition, the collection covers almost every
aspect on PAT from mathematics, image reconstruction methods, instrumentation,
and phantom/animal experiments to clinical applications. I believe this book will be
particularly useful for graduate students and researchers who wish to enter the field
of PAT.
In Chapter 1, the fundamentals of PAT are presented including the theory on photo-
acoustic effect, various image reconstruction methods, and instrumentation. Chapter
2 describes various advanced methods for quantitative PAT, which allow the recov-
ery of tissue optical absorption coefficients and/or acoustic properties. Experimental
validations of the reconstruction methods are also presented here. The develop-
ment of several image-enhancing schemes including both software and hardware
approaches are discussed in Chapter 3. Chapter 4 describes array-based PAT systems
that are the foundation for the realization of 2D, 3D, and 4-D PAT. Photoacoustic
microscopy (PAM), a variation of PAT, is detailed in Chapter 5. The recent devel-
opment of multimodality methods is presented in Chapter 6. These works describe
the combinations of PAT/PAM with other imaging methods including DOT, fluores-
cence molecular tomography (FMT), ultrasound imaging and OCT, which represent
new directions in the field of PAT. Chapter 7 focuses on the discussion of contrast
agents–based molecular PAT where both nontargeted and cell receptor–targeted
methods are included. Chapter 8, the final chapter, describes the clinical applications
and animal studies in the areas of breast cancer detection, osteoarthritis diagnosis,
seizure localization, intravascular imaging, and image-guided cancer therapy.
I wish to thank my previous and current colleagues whose contributions to the
works presented have made this book possible. I would also like to thank the staff
at CRC Press, especially associate editor Ashley Gasque and project coordinator Jill
Jurgensen, who have worked diligently at bringing this book to fruition. Finally, my
wife, Yonghong, and my family made it all worthwhile.
Huabei Jiang
Gainesville, Florida
About the Author
Huabei Jiang, PhD, is J. Crayton Pruitt Family professor in the Department of
Biomedical Engineering at the University of Florida. He has published more
than 300 peer-reviewed scientific articles and patents. Dr. Jiang is a fellow of the
Optical Society of America (OSA), a fellow of the International Society of Optical
Engineering (SPIE), and of the American Institute of Medical and Biological
Engineering (AIMBE).
xiii
1 Fundamentals of
Photoacoustic
Tomography
Photoacoustic tomography (PAT), or optoacoustic tomography (OAT), is based

on the photoacoustic (PA) effect discovered by Alexander Graham Bell (1880),
who observed that audible sound is produced when chopped sunlight is incident
on optically absorbing materials. Since then, techniques based on PA spectros-
copy and microscopy have been developed and used primarily for the studies of
gases and solids (Tam 1986). Methods for tomographic PA imaging or PAT did
not appear until 1990s first for nonbiological medium (Gusev and Karabutov
1993) and then for biological tissue (Oraevsky et al. 1994, Kruger and Liu 1994).
PAT is basically concerned with an inverse problem where a single short-pulsed
light beam illuminates an object and the photoacoustic waves excited by ther-
moelastic expansion are measured using wideband ultrasound transducers in
multiple locations around the object. The geometry of the object and spatial
distribution of the optical absorption are obtained from the measured scattered
fields using a reconstruction algorithm or formed directly by time-resolved pho-
toacoustic signals. In this chapter, we present the basic PA theory, several image
reconstruction methods, and essential hardware components needed for imple-
menting a PAT system.
1.1 PHOTOACOUSTIC EFFECT
1.1.1 Empirical Description
The basic physical principles behind PA effect can be demonstrated using some
approximate solutions. For generality, a 3D model using lateral dimensions x and y and
depth z needs to be considered. Biological tissue is homogenously irradiated with a
short laser pulse of energy E (J) on an area A (m2), and the photon density or radiant
exposure (J/m2) on the surface can be approximately stated as,
Φ = Φ (x, y, 0) = E/A (1.1)
The radiant exposure for a large depth z is estimated by,
I ( x , y, z ) ≈ Φ 0 e − eff z (1.2)

1
2 Photoacoustic Tomography
where μeff is the effective optical attenuation coefficient, which is a function of opti-
cal absorption coefficient, a and reduced scattering coefficient, s ,
eff = 3 a ( a + s ) (1.3)

It should be noted that the photon density here can be accurately calculated
using the photon diffusion equation with any kind of source distribution. In PA
effect, the stress generation depends on the locally absorbed light energy density,
Ψ that is the product of optical absorption coefficient, a and photon density, Φ
(Norton and Vo-Dinh 2003),
Ψ= Φ (1.4)
a
With mass density, ρ (kg/m3) and specific heat, Cp (J/(kg°C)), the absorbed light
energy leads to a temperature increase (°C),
T ( x , y, z ) = Ψ( x , y, z )/(ρ.C p ) (1.5)

Additionally due to laser excitation, the absorbed light energy also gives rise to a
fractional volume expansion dV/V,
dV/V = −κp + βΔT (1.6)
where κ denotes the isothermal compressibility, β is the thermal coefficient of vol-

ume expansion, and ΔT and p are the changes in temperature (k) and pressure (Pa),
respectively. The isothermal compressibility, κ, can be expressed as,
κ = C p /(ρvs2Cv ) (1.7)

where Cp and Cv are, respectively, the specific heat capacities at constant pressure
and volume, and vs is the acoustic velocity. If the laser excitation is in both the ther-
mal and stress confinements, the fractional volume expansion is negligible and the
local pressure rise can be obtained from Equation (1.6),
p0 = β T/κ (1.8)

If we define a set of coefficients as Grueneisen parameter, Γ (Gusev and

Karabutov 1993),
Γ = β/(κρCV ) = βvs2 /C p (1.9)

Equation (1.8) is rewritten in consideration of Equations (1.4), (1.5), and (1.9),
p0 = β a Φ/κρCv = Γ a Φ (1.10)

Fundamentals of Photoacoustic Tomography 3
For water and diluted aqueous solution, Γ can be estimated by the following
empirical equation,
Γ (T0 ) = 0.0043 + 0.0053T0 (1.11)
in which T0 is the temperature in degrees Celsius. p 0 is often specified as the initial

pressure for a temporal PA wave equation. These sets of empirical equations for
the calculation of p 0 should satisfy the stress confinement and thermal confinement
conditions (Gusev and Karabutov 1993, Kostli et al. 2000). The direct efficient con-
version to pressure is possible if the acoustic transient does not leave the heated
region during the heating process. This condition is called stress confinement and is
satisfied if the pulse duration tp is smaller than the propagation time of the acoustic
transient through the region d with acoustic velocity vs (Gusev and Karabutov 1993,
Kollkman et al. 2007),
τ p < d /vs (1.12)

If the laser pulse width is much shorter than the thermal relaxation time, τth , the
excitation is said to be in thermal confinement, and heat conduction is negligible dur-
ing the laser excitation. The thermal confinement condition is written,
τth < d 2 /α th (1.13)
in which α th is the thermal diffusivity (m2/s) and dc ⊕1/ eff .
Example
A laser pulse with energy E = 30 mJ, pulse length = 10 ns, beam radius r = 1 cm
and wavelength = 550 nm irradiates human blood vessel. The radiant exposure is,
Φ0 = E/πr2= 30 mJ/3.14 cm2 = 9.55 mJ/cm2
Assuming μa = 25 cm –1, the absorbed light energy density is calculated as,
Ψ = 25 × 9.55 = 238.75 mJ/cm3 = 2.38 × 105 J/m3
Then we can obtain the temperature increase and initial pressure (assuming ρ =
1 g/cm3 and Cp= 4 Jg–1K–1),
T ( x , y , z ) = Ψ( x , y , z )/(ρC p ) = 238.75 mJ/cm3 /(1g/cm3 × 4 Jg −1K −1) = 59.7 mk

p0 = ΓΨ = 0.2 × 2.38 × 105 Pa = 0.2 × 2.38 bar = 0.476 bar

To obtain this pressure, stress confinement with a region d = 1/μa = 400 μm and
sound velocity vs = 1.5 μm/ns needs to be satisfied: τp = 10 ns < d/vs = 266 ns.
1.1.2 Rigorous Theory
The empirical description presented in Section 1.1.1 is intended to illustrate the PA
effect and certainly cannot be used for the purpose of imaging. The PA wave genera-
tion and propagation in biological tissue is rigorously described by the Helmholtz-
like wave equation. To derive the PA wave equation, three general elastic mechanical
and fluid dynamic equations are responsible for the PA generation and propagation:
Newton’s law of motion, equation of continuity, and thermal elastic equation:
∂
ρ(r ) V (r , t ) = − p(r , t ) (1.14)
∂t
1 ∂ ∂
⋅ V (r , t ) = − 2
p(r , t ) + β T (r , t ) (1.15)
ρ(r ) vs (r ) ∂t ∂t
∂
ρ(r )C p T (r , t ) = H (r , t ) (1.16)
∂t
where V is particle velocity and H is the laser excitation source term which can be
written as H = ΨI(t), where I(t) is the temporal illumination function. Considering
Equations (1.14)–(1.16) and eliminating V, we obtain
1 1 ∂2 β ∂
ρ(r ) ⋅ p(r , t ) − 2 2
p(r , t ) = − H (r , t ) (1.17)
ρ(r ) vs (r ) ∂t C p ∂t

If a homogeneous elastic medium is assumed, Equation (1.17) is further written,
2 1 ∂2 β ∂
p(r , t ) − p(r , t ) = − H (r , t ) (1.18)
2
v (r ) ∂t 2
s C p ∂t

Equation (1.18) is the general PA wave equation (without acoustic attenuation) in

time-domain. If a homogeneous acoustic medium is also assumed, Equation (1.18)
is written as
2 ∂2 β ∂
p(r , t ) − p(r , t ) = − H (r , t ) (1.19)
v ∂t 2
2
0 C p ∂t

where v0 is the constant acoustic velocity within the entire problem domain. Denoting
the following Fourier transform expression for acoustic pressure,
+∞

P(r , ω ) =
∫ p(r , t) exp(−iωt) dt (1.20)
−∞
and taking the Fourier transform on Equation (1.18) with respect to variable t, we obtain
the PA wave equation in frequency-domain (assuming H = ΨI (t ) = Ψδ(t − t0 ) ):
2 βΨ(r )
P(r , ω ) + k 2 P(r , ω ) = ik vs (1.21)
Cp

where P is the pressure wave in frequency-domain; k = ω/vs is the wave number

described by the angular frequency, ω and the acoustic speed, vs. The PA wave
equation in frequency-domain for a homogeneous acoustic medium is accordingly
expressed as
2 βΨ(r )
P(r , ω ) + k02 P(r , ω ) = ik0 v0 (1.22)
Cp

where k0 = ω/v0 . If both the acoustic velocity and attenuation are heterogeneously
distributed, the frequency-domain PA wave equation is then described as:
2 βΨ(r )
P(r , ω ) + k02 (1 + O) P(r , ω ) = ik0 v0 (1.23)
Cp

where O is a coefficient that depends on both acoustic velocity and attenuation as follows:
O = vo2 /vs2 − 1 + iAv0 /k0 vs2 (1.24)

where v0 is the constant acoustic velocity in a reference/coupling medium, and A is

the acoustic attenuation coefficient.
1.2 IMAGE RECONSTRUCTION METHODS

1.2.1 Delay-and-Sum Beam Forming Algorithm
To form the PA image of an object, an image reconstruction algorithm is required.
A simple, yet very effective, method is the delay-and-sum beam forming algorithm
that is commonly used in radar signal processing (Hoelen and de Mul, 2000)). The
image expression in the case of near-field for photoacoustic imaging can be stated as
∑ w S (t + δ )
i
i i i
S (r , t ) = (1.25)
∑w i
where S(r, t) is the image output at a particular spatial point r within the imaging
domain, Si(t) is the time-resolved acoustic signal from the ith receiver, δi is the
delay applied to this signal, and wi is an amplitude weighting factor, which is used
to enhance the beam shape, to reduce sidelobe effects, or to minimize the noise
level. Note that δi = |Ri – r|/v is just the time needed for the acoustic signal to travel
from point r to point Ri, where Ri is the spatial location of the ith receiver and v is
the assumed constant acoustic velocity. The summed signal is typically normalized
to make the output independent of the actual set of transducers. This delay-and-sum
algorithm has been tested and evaluated using considerable phantom and in vivo
experiments (Hoelen and de Mul 2000, Manohar et al. 2007).
1.2.2 Iterative Nonlinear Algorithm

While the delay-and-sum algorithm described above is simple to implement, it does
not account for the diffraction effect that is essential to the PA waves. In particular, a
major assumption has to be made in this and other linear algorithms such as backpro-
jection or filtered backprojection seen in the literature (e.g., backprojection ([Kruger
et al. 1995, Wang et al. 2007]) that biological tissues are acoustically homogeneous,
which may not be true in reality). This assumption will certainly affect the accuracy
in image reconstruction. In addition, acoustic properties cannot be reconstructed
because of this assumption. Importantly, functional parameters such as deoxy- and
oxy-hemoglobin (HbR and HbO2) and water content (H2O) cannot be obtained with
these linear methods (see more about quantitative PAT in Chapter 2).
To overcome the above-mentioned limitations, here we describe a nonlinear
reconstruction algorithm based on the finite element (FE) solution to the full PA
wave equation without the homogeneous acoustic property assumption made thus
far in the literature. Finite element method (FEM) has been a powerful numerical
method for solving the Helmholtz-type equation, because of its computational effi-
ciency and unrivaled ability to accommodate tissue heterogeneity and geometrical
irregularity as well as allow complex boundary conditions and source representa-
tions. This reconstruction approach in PAT is an iterative Newton method with
combined Marquardt and Tikhonov regularizations that can provide stable inverse
solutions. The approach uses the hybrid regularizations-based Newton method to
update an initial optical property distribution iteratively in order to minimize an
object function composed of a weighted sum of the squared difference between
computed and measured data. Together with the iterative Newton method, this
nonlinear algorithm is able to precisely solve the Helmholtz wave equation and
fulfill reliable inverse computation for an arbitrary measurement configuration.
The nonlinear algorithm is implemented in both frequency- and time-domain as
follows.
1.2.2.1 Frequency-Domain FE-Based Algorithm

The PA wave equation in frequency-domain is written if a homogeneous acoustic
medium is considered (for the case of inhomogeneous acoustic medium, see Chapter 2),
2 βΨ(r ) (1.26)
P(r , ω ) + k02 P(r , ω ) = ik0 v0
Cp
Expanding acoustic pressure, P as the sum of coefficients multiplied by a set of

basis functions, ψj: P = ∑ Pj ψ j , the finite element discretization of the Helmholtz
wave equation (1.26) can be expressed as
N
∂2 ψ j
∑j =1
Pj ψ j ⋅ ψ i − k02 ψ j ψ i −
∫ ηψ j + γ
∂φ2
ψ i ds
= − ik0 v0βΨ/CP ψ i (1.27)

where the following second-order absorbing boundary conditions have been applied
(Jiang et al. 2006):
∂2 P
P ⋅ nˆ = ηP + γ (1.28)
∂φ2
where η = (−ik0 − 3/2ρ + i3/8 k0 ρ2 )/(1 − i/k0 ρ) and γ = (−i/2 k0 ρ2 )/(1 − i/k0 ρ) , N is the
total number of nodes of the finite element mesh, <(⋅)> indicates the integration over
the problem domain, and ≡ expresses the integration over the boundary. In both the
forward and inverse calculations, the unknown Ψ needs to be separated into real (ΨR)
and imaginary (ΨI) parts, both of which are expanded in a similar fashion to P as a
sum of unknown parameters multiplied by a known spatially varying basis function.
The matrix form of Equation (1.27) is expressed as follows:
[ A]{P } = {B} (1.29)
where
∂2 ψ j

Aij = ψ j ⋅ ψ i − k02 ψ j ψ i −
∫ ηψ j + γ
∂φ2
ψ i ds (1.30)
Bi = − Ik0 v0β ∑Ψ
k
R,k ψ k ψi C p + k 0 v0 β ∑Ψ
l
I ,l ψlψi C p (1.31)

{P} = {P1 , P2 ,…, PN }T (1.32)
To form an image from a presumably uniform initial guess of the optical prop-
erty distribution, we use iterative Newton’s method to update ΨR and ΨI from their
starting values. In this method, we Taylor expand P about an assumed (ΨR, ΨI)
R ,Ψ
distribution, which is a perturbation away from some other distribution, (Ψ I ),
such that a discrete set of P values can be expressed as
P (Ψ I ) = P (Ψ R , Ψ I ) + ∂ P Ψ R + ∂ P
R ,Ψ Ψ I + (1.33)
∂Ψ R ∂Ψ I
where Ψ R = Ψ R − Ψ R , and Ψ I = Ψ I − Ψ I . If the assumed optical property dis-

tribution is close to the true profile, the left-hand side of Equation (1.33) can be
considered as true data (observed or measured), and the relationship can be truncated
to yield
J χ = P o − P c (1.34)

where
∂ P1 /∂Ψ R ,1 ∂ P1 /∂Ψ R , K ∂ P1 /∂Ψ I ,1 ∂ P1 /∂Ψ I , L

∂ P2 /∂Ψ R ,1 ∂ P2 /∂Ψ R , K ∂ P2 /∂Ψ I ,1 ∂ P2 /∂Ψ I , L
J= (1.35)

∂ PM /∂Ψ R ,1 ∂ PM /∂Ψ R ,K ∂ PM /∂Ψ I ,1 ∂ PM /∂Ψ I , L
χ = ( Ψ R ,1 , Ψ R ,2 ,…, Ψ R ,K , Ψ I ,1 , Ψ I ,2 ,…, Ψ I , L )T (1.36)

P o = ( P1o , P2o ,…, PMo )T (1.37)
P c = ( P1c , P2c ,…, PMc )T (1.38)
and Pi o and Pic are measured and calculated acoustic field data based on the assumed
(ΨR, ΨI ) distribution data for i = 1,2,…M boundary locations. ΨR,k (k = 1,2, …, K)
and ΨI ,l (l = 1,2,…, L) are the reconstruction parameters for the optical property
profile. In order to realize an invertible system of equations for χ, Equation (1.34)
is left multiplied by the transpose of J to produce
( J T J + λI) χ = J T (P o − P c ) (1.39)

where regularization schemes are invoked to stabilize the decomposition of J T J;
I is the identity matrix, and λ is the regularization parameter determined by com-
bined Marquardt and Tikhonov regularization schemes. We have found that when
T
λ = (P o − P c ) × trace [ J J] , the reconstruction algorithm generates best results for
PAT image reconstruction. The process now involves determining the calculated
acoustic field data and Jacobian matrix. The reconstruction algorithm here uses the
hybrid regularization-based Newton method to update an initial (guess) optical prop-
erty distribution iteratively via the solution of Equations (1.29) and (1.39) so that an
object function composed of a weighted sum of the squared difference between com-
puted and measured acoustic pressures for all acoustic frequencies and optical wave-
lengths can be minimized.
1.2.2.2 Time-Domain FE-Based Algorithm

The time-domain PA wave equation (1.19) for an acoustically homogeneous medium
can be rewritten as
2 1 ∂2 p(r, t ) βΨ (r) ∂ I (t )
p(r, t ) − 2 2
=− (1.40)
v0 ∂t Cp ∂t

where I (t ) = δ(t − t0 ) is typically assumed.

Expanding p as the sum of coefficients multiplied by a set of basis function ψj,

p = ∑ ψ j p j , the finite-element discretization of Equation (1.40) is expressed as
N N
∂2 p j 1
∑ ∫
j =1
pj
S
ψ i ⋅ ψ j dS + ∑ j =1
∂2 t ∫ S v02
ψ i ψ j dS −
∫ ψ
l
i p ⋅ nˆ dl
βΨ ∂ I

=
∫ S C P ∂t
ψ i dS (1.41)
The Bayliss-Turkel radiation boundary conditions are employed here (Jin

2002),
Bm p = 0 (1.42)

where Bm is a sequence of linear differential operators,
m
1 ∂ ∂ 4l − 3
Bm = ∏ +
v0 ∂t ∂r
+
2r
(1.43)
l =1
and Bm is subject to the accuracy,
1
Bm = O 1
(1.44)
2 m +1+
r 2
This method provides an even more accurate approximation to the Sommerfeld

radiation condition by annihilating higher order terms in 1/r. However, the higher
order (i.e., m > 2) boundary operators are not recommended for numerical imple-
mentation since they tend to spoil the sparsity of the finite element matrix. As such,
the first order absorbing boundary operator method is described as
∂p 1 ∂p p
B1 p = + + = 0 (1.45)
∂r v0 ∂t 2r
The boundary term based on Equation (1.42) is rewritten
1 ∂p p
p ⋅ nˆ = − − . (1.46)
v0 ∂t 2r
In both the forward and the inverse calculations, the unknown coefficients Ψ need
to be expanded in a similar fashion to p as a sum of unknown parameters multiplied
by a known spatially varying basis function. Thus, Equation (1.41) can be expressed
as the following matrix form in consideration of Equation (1.46),
[ K ]{p} + [C ]{p} + [ M ]{

p} = [ B] (1.47)

where the elements of the matrix are written
1
K ij =
∫ ψ i ⋅ ψ j dS +
2r ∫ ψ ψ dl (1.48)
l
i j
1

Cij =
v0 ∫ ψ ψ dl (1.49)
l
i j
1

M ij =
v02 ∫ ψ ψ dS (1.50)
S
i j
β ∂Ι
Bi =
Cp ∫ ψ ∑ψ Ψ
S
i
k
k k dS ⋅
∂t
(1.51)

{p} = {p1 , p2 ,…, pN }T (1.52)

T
∂ p1 ∂ p2 ∂p
{p} = , ,…, N (1.53)
∂t ∂t ∂t
T
∂2 p1 ∂2 p2 ∂2 p
{
p} = 2
, 2 ,…, 2 N (1.54)
∂ t ∂ t ∂ t

Note that here Newmark’s time-stepping scheme has been used for the discreti-
zation of time dimension (Smith and Griffiths 2004). The Newmark algorithm is a
commonly used implicit method for the second order propagation equations such as
Equation (1.47). In our case, we have
{P}t + t = {P}t + [(1 − δ){P}t + δ{P}t + t ] t (1.55)
1
{P}t + t = {P}t + {P}t t + − α {P}t + α{P}t + t t 2 (1.56)
2
where Δt is the time interval; α and δ are the time-stepping parameters, which deter-
mine the accuracy and stability of the algorithm; {P}t ,{P}t , and {P}t are the values
of the pressure and its derivatives at time t; and {P}t + t ,{P}t + t , and {P}t + t are the
values of the pressure and its derivatives at the subsequent instant t + Δt.
Assuming that [ K ],[C ],[ M ], α, δ and Δt are constant, and from Equations (1.47),
(1.55), and (1.56), we obtain the following formula for calculating {P}t + t from
{P}t ,{P}t , and {P}t :
1 δ 1 1 1
[K ] + [M ] + [C ] {P}t + t = [ B]t + t + [ M ] {P}t + {P}t + − 1 {P}t
α t2 α t α t2 α t 2α
δ δ δ
+ [C ] {P}t + − 1 {P}t + −1 t{P}t
α t α 2α

(1.57)
The derivatives of the pressure at the subsequent instant are computed as follows:
1 1 1
{P}t + t = 2
({P}t + t − {P}t ) − {P}t − − 1 {P}t , (1.58)
α t α t 2α
{P}t + t = {P}t + t (1 − δ){P}t + δ t{P}t + t (1.59)
δ = 0.5 and α = 0.25 are typically used in the image reconstruction. When
δ ≥ 0.5, and α ≥ 0.25(0.5 + δ)2, the Newmark algorithm preserves unconditional
stability, which means that the amount of Δt does not affect the stability of the solu-
tion to the equation (Bath 1976).
Following the procedures described in Section 1.2.2.1, the matrix equation for
inversion can be written as

(J T
t Jt + λI ) χ = JtT ( Pt o − Ptc ) . (1.60)
∂P
where Jt is the time-dependent Jacobian matrix formed by ∂Ψ at each time step,
o ,c
Pt is the time-dependent measured and calculated acoustic field data for i =
1,2,… M boundary locations, χ = ( Ψ1 , Ψ 2 ,… Ψ N )T is the updating vector,
I is the identity matrix, and λ is the regularization parameter determined by com-
bined Marquardt and Tikhonov regularization schemes. Here the submatrix of the
Jacobian for each time step is assembled to form the full Jacobian matrix for all time
steps. Thus we just need to solve Equation (1.47) one time per iteration for the whole
time range of interest.
1.2.3 A-Line/B-Mode Image Formation Method

The image reconstruction methods described in Sections 1.2.1 and 1.2.2 are usually
used for PAT based on unfocused ultrasound transducer(s) for large tissue imaging.
One form of PAT is termed photoacoustic microscopy (PAM), which uses a focused
high-frequency transducer for superficial imaging typically in reflection mode (see
Chapter 5 for more discussion on PAM). Since PAM uses focused transducer, its
image can be simply formed by the A-line or B-mode image formation method
used in conventional ultrasound imaging. In this method, one-dimensional depth-
resolved images (i.e., A-line) at each transducer location are acquired and additional
raster scanning of the transducer along a transverse direction (x or y) produces the
2D images referred to as B-scan. Further raster scanning along the other transverse
direction enables the reconstruction of a 3D image. The photoacoustic images can
be displayed through cross-sectional (B-scan) images, or maximum amplitude pro-
jection (MAP) images, where a MAP image is formed by projecting the maximum
photoacoustic amplitudes along a direction (typically z) to its orthogonal plane (x–y).
Let us use an example to illustrate how one A-line image is formed. A target (hair)
is embedded in a background phantom (Figure 1.1a). Figure 1.1b presents an A-line
acoustic signal collected along z-axis crossing the center of the hair target. The depth
of an image pixel (e.g., L for the center of the hair target shown in Figure 1.1b) is
obtained by multiplying the delay time or time difference between the launching
of the laser pulse and the receiving of the acoustic signal by the assumed constant
acoustic speed in phantom/tissue (typically 1.5 mm/μs). In order to achieve the best
spatial resolution, the target should be located at the focal plane of the transducer.
Figure 1.1c gives the recovered 2D photoacoustic image of the hair target along x-z
plane.
x
Transducer
z
Water 0
0.5
Surface 1
1.5
2
L
2.5
3
Phantom Target
3.5
4
4.5
5
5.5
0.05 0 0.05 0.1
Amplitude (a.u.)
(a) (b) (c)
FIGURE 1.1 (a) Schematic showing a hair target embedded in a phantom. (b) A-line acous-
tic signal collected along z-axis crossing the center of target (dashed line along z-axis in (a)).
(c) Recovered 2D image of the hair target along x-z plane.
Trigger
Reflecting
mirror Pulsed laser
PC
Concave lens
MCA
Rotator
RH LH
Rotator Water tank

controller
ROB LOB
Membrane
Pulser & 2 mm
receiver
Transducer 0 0.3 0.6 1
FIGURE 1.2 Schematic of a typical PAT system. The sample here is a rat. The membrane is
used to isolate the water and rat while ensuring the target (i.e., rat head) is within the imaging
domain. The insert shows one slice of tomographic photoacoustic image of the rate brain. PC:
personal computer.
1.3 INSTRUMENTATION
A typical PAT system consists of a short-pulsed laser (e.g., several nanoseconds), a
singe transducer scanning subsystem, and acoustic signal detection subsystem (see
Figure 1.2). In this system, pulsed light from a Nd:YAG, Ti:Sapphire, or diode laser
is coupled into the sample via an optical subsystem and generates acoustic pressure
wave. A transducer (1 MHz or higher central frequency) is used to receive the acous-
tic signals. The transducer and the sample are immersed in a water tank. A rotary
stage rotates the receiver relative to the center of the tank. One set of data is taken at
multiple receiver positions (e.g., 120 positions) when the receiver is scanned circu-
larly over 360°. The complex wavefield signal is first amplified by a preamplifier and
then amplified further by a pulser/receiver. A data acquisition board converts it into
digital one, which is fed to a computer. The entire data acquisition can be realized
through C programming, for example. In this system, data collection for a total of
120 measurements may require up to 2 minutes.
In the typical PAT setup shown, the mechanical scanning significantly reduces
the speed of data acquisition. The use of a transducer array has clear acquisition time
advantages over a mechanically scanned system, making real-time data acquisition
possible. In addition, a transducer array-based system eliminates the use of a water
tank and allows the direct contact of the array with the skin (of course ultrasound
gel needs to be applied as coupling medium). The major drawback of an array-based
system, however, is its significantly added cost. We will detail an array-based PAT
system in Chapter 4.
The light illumination pattern can be an area or multiple points depending on

the applications—the key here is that the tissue needs to be illuminated homo-
geneously with sufficient optical energy being absorbed. In addition to the cir-
cular scanning of a single transducer or an arc or circular array of transducers,
other types of transducer arrangements are possible. For example, reflection- or
backward-mode laser excitation/ultrasound detection is commonly used. In such
configurations, the laser excitation and ultrasound detection are performed at the
same side, which is very convenient for applications such as intraoperative and skin
imaging (see Chapter 5).
2 Quantitative
Photoacoustic
Tomography
While the image reconstruction methods presented in Chapter 1 and the linear algo-
rithms such as backprojection and filtered backprojection seen in the literature can be
used to provide high-quality photoacoustic (PA) images, the PA images provided are
qualitative in nature as they represent only the distribution of absorbed light energy
density that is the product of the intrinsic optical absorption coefficient and extrinsic
optical fluence distribution, indicating that the absorbed energy density as an imag-
ing parameter of conventional photoacoustic tomography (PAT) is clearly not an
intrinsic property of tissue. It is well known, however, that it is the tissue absorption
coefficient that directly correlates with tissue physiological/functional information.
These physiological parameters including hemoglobin concentration, blood oxygen-
ation, and water content are critical for accurate diagnostic decision making.
While the term quantitative PAT means primarily the ability to recover the opti-
cal absorption coefficient distribution, we show in this chapter that quantitative PAT
can also extract tissue acoustic property distribution such as acoustic velocity and/or
attenuation. It is clear that neither delay-and-sum method nor linear algorithms such
as backprojection/filtered backprojection can do so. Fortunately, the iterative finite
element (FE)-based nonlinear algorithms described in Chapter 1 can be extended
to achieve these quantification abilities. Since the first report of quantitative PAT
in 2006 (Yuan and Jiang 2006), several quantitative PAT methods have been devel-
oped and experimentally validated (Yuan et al. 2006, 2007; Yin et al. 2007; Yao
et al. 2009; Yuan and Jiang 2012). Here we describe four methods for recovering
absorption coefficient and one method for simultaneous reconstruction of optical
and acoustic properties, all FE-based in frequency-domain. We use phantom experi-
ments to demonstrate the validation of each quantitative method.
2.1 RECOVERY OF OPTICAL ABSORPTION

COEFFICIENT: METHOD 1
Method 1 for quantitative PAT includes two steps. The first is to obtain the map of
absorbed optical energy density through the FE-based iterative reconstruction algo-
rithm described in Chapter 1. The second step is to recover the distribution of opti-
cal absorption coefficient from the absorbed energy density obtained from the first
step based on the finite element solution to the photon diffusion equation. To ensure
the reconstructed absorbed energy density in absolute units, a calibration with an
15
absorption-dominated acoustic phantom needs to be performed in order to obtain a

constant, ξ , that represents the relationship between the absolute pressure preal and
the measured temporal acoustic signals pc . preal at the surface of the laser-irradiated
phantom/tissue can be approximately written (Yuan and Jiang 2006),
preal = Γ a Φ 0 = ξpc (2.1)

in which Γ is the Grüneisen coefficient (Γ = c0β/Cp), and Φ0 is the incident laser flu-
ence at the surface of the phantom/tissue. Once ξ in Equation (2.1) is obtained, the
absolute absorbed energy density for PAT imaging can be obtained from the absolute
acoustic pressure that is the product of ξ and measured acoustic signal. In addition,
the nonlinear relationship between the optical fluence within the tissue and optical
absorption coefficient can be approximated by the photon diffusion equation and
type III Boundary Conditions (BCs) BCs as follows (Jiang 2010),
⋅ D(r ) Φ(r ) − α (r )Φ(r ) = − S (r ), − D Φ ⋅ nˆ = αΦ (2.2)

in which Φ is the photon density or optical fluence, and D is the diffusion coefficient
and can be written as D = 1/(3( a + s )), where s is the reduced scattering coef-
ficient. α is a coefficient related to the internal reflection at the boundary. S is the
distributed source due to pulsed excitation and S = Ea / ∫ V1 dV1, in which Ea is the
absorbed pulsed optical energy. V1 is the optically absorbed volume integrated by
illumination area over penetration depth across subsurface. As such, if incident laser
source, absolute absorbed energy density Ψ, and reduced scattering coefficient dis-
tribution are estimated in advance, μa can be determined by the following iterative
solution procedure using finite element method (Yuan and Jiang 2006): (a) choose an
initial value for μa, e.g., μa= 0.0001 mm–1; (b) compute the optical fluence Φ using
the diffusion equation; (c) calculate the real part of absorbed energy density by Ψc =
μa Φ; (d) compute the error between the real part of absolute Ψ and calculated Ψc,
and μa is updated using μa = Ψ/Φ; (e) if the error is sufficiently small, then the itera-
tive calculation stops; otherwise repeat steps (b)–(d) until a small error is reached.
To reduce the boundary noise effect, the absorption coefficient within the boundary
domain needs to be assumed to have the same value as the background medium.
Examples
Experiments were performed to validate Method 1 described above. The experi-

mental system used is similar to the one depicted in Figure 1.1. Briefly, pulsed
light from an Nd: YAG laser (wavelength: 532 nm, pulse duration: 3–6 ns) was
coupled into the phantom via an optical subsystem and generated acoustic sig-
nals. The transducer and phantom were immersed in a water tank. A rotary stage
rotated the receiver relative to the center of the tank. A 1 MHz transducer was
used to receive the acoustic signals. The incident fluence was controlled below
10 mJ/cm2 and the incident laser beam diameter was 3 cm. In the first three exper-
iments, we embedded two nanoparticle-containing objects (3 mm in diameter) in
a 50 mm-diameter solid cylindrical phantom where the two objects had different
Quantitative Photoacoustic Tomography 17
optical contrasts relative to the phantom. We then immersed the object-bearing

solid phantom into a 110.6 mm-diameter water background (height: 3 cm). The
absorption coefficient of the phantom was 0.005/mm. In the fourth experiment,
the phantom materials used consisted of Intralipid® as scatterer and India ink as
absorber with Agar powder (1–2%) for solidifying the Intralipid and India ink solu-
tion. Two objects (5.5 mm in diameter) were embedded into a 50 mm-diameter
solid cylindrical phantom. The absorption coefficient of the phantom and targets
were 0.01/mm and 0.07/mm, respectively.
Figures 2.1a–c present the reconstructed optical absorption images of two
objects having 0.5, 0.25, and 0.125 nM nanoparticles in each of the objects,
respectively. We see that the object(s) in each case are clearly detected. By esti-
mating the full width at half maximum (FWHM) of the optical absorption property
profiles shown in Figure 2.2, we found that the recovered object size ranged from
30
0.05
20 0.045
0.04
10 0.035
0.03
0
0.025
0.02
10
0.015
20 0.01
0.005
30
30 20 10 0 10 20 30
(a)
30
0.018
20 0.016
0.014
10
0.012
0 0.01
0.008
10
0.006
0.004
20
0.002
30
30 20 10 0 10 20 30
(b)
FIGURE 2.1 Reconstructed optical absorption coefficient images (units: mm–1). (a), (b), and (c):
Two targets containing 0.5, 0.25, and 0.125 nM nanoparticles in each target, respectively. (continued)
× 103
30 8
7
20
6
10
5
0 4
3
10
2
20
1
30
30 20 10 0 10 20 30
(c)
FIGURE 2.1 (continued) Reconstructed optical absorption coefficient images (units: mm–1).
(a), (b), and (c): Two targets containing 0.5, 0.25, and 0.125 nM nanoparticles in each target,
respectively.
Absorption Coefficient (mm–1)
0.050 0.020
0.045 0.018
0.040 0.016
0.035 0.014
0.030 0.012
0.025 0.010
0.008
0.020 0.006
0.015 0.004
0.010 0.002
0.005 0.000
–4 –2 0 2 4 –6 –4 –2 0 2
x (mm) y (mm)
(a) (b)
0.008
0.007
0.006
0.005
0.004
0.003
0.002
–4 –2 0 2 4
x (mm)
(b)
FIGURE 2.2 Recovered optical absorption profiles plotted along y = 0 mm (a), x = 0.5 mm
(b), and y = 0 mm (c) from the images shown in Figures 2.1(a)–(c), respectively.
2.8 to 3.3 mm, which is in good agreement with the actual object size of 3 mm.
Moreover, the nanoparticles concentration used in our experiments (0.125, 0.25,
and 0.5 nM) gives an absorption coefficient of the object of 0.01, 0.025, and
0.05 mm–1, respectively. As can be seen from Figures 2.1 and 2.2, we are able to
obtain quantitatively resolved images in terms of the size as well as absorption
coefficient of the objects.
The reconstructed absorption coefficient image for the fourth experiment is
shown in Figure 2.3a. The recovered object size in this case was calculated to
be 5.0 mm, and the peak value of recovered absorption property is 0.072/mm,
as plotted in Figure 2.3b. Again, both values are in good agreement with the
exact values.
We also note that the absorption coefficient value of the objects is more accu-
rately recovered for the higher contrast cases (see Figures 2.1a and 2.3a) than that
for the lower contrast cases (Figures 2.1b and c). The error between the recon-
structed and exact absorption coefficient is up to 31% for the low-contrast cases.
The degradation is most likely due to the limitations associated with the simple
20
0.07
15
0.06
10
0.05
5
0.04
0
5 0.03
10 0.02
15 0.01
20
20 10 0 10 20
(a)
0.08
0.07
0.06
0.05
0.04
0.03
0.02
0.01
0.00
–0.01
–4 –2 0 2 4 6 8
x (mm)
(b)
FIGURE 2.3 (a) Recovered optical absorption coefficient image (a) (units: mm–1), (b)
Optical absorption profile along y = –5.5 mm from the image shown (a).
iteration procedure for calculating μa, because we have to know the strength
of incident light in advance, perform careful calibration with the nonscattering
medium, and assume the reduced scattering coefficient of the phantom known
as a constant (1.0 mm–1 for the first three cases and 3.0 mm–1 for the fourth case).
We will see that many of these limitations can be overcome using the methods
described in the following three sections.

Method 2 for quantitative PAT is aimed to overcome the limitations associated with
Method 1 presented in Section 2.1. Specifically, this method combines the conven-
tional PAT with diffusing light measurements coupled with an optimization proce-
dure based on the photon diffusion equation. Method 2 also includes two steps. The
first is the same as Method 1, i.e., to obtain the map of absorbed optical energy den-
sity through the frequency-domain FE-based algorithm described in Chapter 1. The
difference is in the second step, which attempts to estimate the distribution of optical
fluence using the experimentally measured boundary optical fluence data coupled
with the photon diffusion equation-based optimization procedure and to recover the
distribution of optical absorption coefficient from the estimated optical fluence and
the absorbed energy density obtained from the first step.
The second step is based on the iterative solution of the following diffusion equa-
tion and χ2 calculation:
⋅ D Φ(r ) − Φ(r ) = − S (r ) (2.3)

α
∑(Φ
2
χ2 = (m)
i − Φ(i c ) ) (2.4)
i =1
where Φ(i m ) and Φ(i c ) are the measured and calculated optical fluence for i = 1,2,…
M boundary locations. The goal of the second step is to obtain the distribution of
optical fluence within the entire imaging domain through an optimization procedure
based on Equations (2.3) and (2.4). This simple least squares minimization scheme
can be explained as follows: Given a range of values of s , a, and S (always avail-
able empirically), we compute the χ2 error as function of s, a , and S where Φ(i m )
is from the measured optical data and Φ(i c ) is calculated from Equation (2.3). The
rationale of this scheme is based on the argument that the minimum of χ2 corre-
sponds to the effective values of s, a, and S associated with the medium of inter-
est. The desired distribution of optical fluence, Φ, is calculated from Equation (2.3)
with the optimized set of s, a, and S in place. Thus, the final distribution of a is
calculated using a = Ψ/Φ where Ψ is obtained from the conventional PAT (the first
step). We found that this division is computationally stable, because there is no pos-
sibility of having a zero or close to zero value for Φ.
Examples
Phantom experiments were conducted to demonstrate the power of Method 2

using the PAT imaging system shown Figure 1.1 with the addition of a simple light
detection subsystem. Diffusing light was collected along the surface of the phan-
tom using a 2 mm-diameter fiber optic bundle coupled with a 2 GHz bandwidth
high-speed photodetector and recorded by a 2.5 GHz bandwidth digital oscil-
loscope. A computer controlled the scanning of the fiber bundle, and 120 optical
measurements were conducted and used in the calculation.
In the first three experiments, we embedded one or two objects with a size
ranging from 3 to 0.5 mm in a 50 mm-diameter solid cylindrical phantom. The
phantom materials used consisted of Intralipid as scatterer and India ink as
absorber with Agar powder (1–2%) for solidifying the Intralipid and India ink solu-
tion. We then immersed the object-bearing solid phantom into the water tank.
The absorption coefficient of the background phantom was 0.01 mm –1, while
the absorption coefficient of the target(s) was 0.03 mm –1. In the final two experi-
ments, we placed a single-target-containing phantom into the water, aiming to
test the capability of detecting target having different optical contrasts relative to
the background phantom. In these two cases, the targets had an absorption coef-
ficient of 0.02 and 0.015 mm –1, respectively. The reduced scattering coefficients
of the background phantom and targets being used in the phantom were 1.0 and
3.0 mm –1 for the first two experiments, and 1.0 and 2.0 mm –1 for the final two
experiments.
Figures 2.4a–c present the reconstructed optical absorption images of one or
two objects having a size of 1.0 mm (a), 2.0 and 3.0 mm (b), and 0.5 mm (c) in
diameter. We see that the object(s) in each case are clearly detected. By estimat-
ing the FWHM of the envelop of the optical absorption property profiles shown
in Figure 2.5, the recovered object size was found to be 1.1, 1.7, 3.2, and 0.7 mm,
which is in good agreement with the actual object size of 1.0, 2.0, 3.0, and
0.5 mm. We also note from Figure 2.5 that the reconstructed images are quantita-
tive in terms of the recovered absorption coefficient value of the objects.
0.03
20
10 0.02
0
0.01
10
0
20
20 0 20
(a)
FIGURE 2.4 Reconstructed optical absorption coefficient images (units: mm–1): (a) a 1 mm-
diameter target, (b) two targets (2 and 3 mm in diameter, respectively), and (c) a 0.5 mm-
diameter target. (continued)
0.03
20
10 0.02
0
0.01
10
0
20
20 0 20
(b)
0.03
20
0.02
10
0 0.01
10 0
20
0.01
20 0 20
(c)
FIGURE 2.4 (continued) Reconstructed optical absorption coefficient images (units: mm–1):
(a) a 1 mm-diameter target, (b) two targets (2 and 3 mm in diameter, respectively), and (c) a
0.5 mm-diameter target.
0.04
0.5 mm
Absorption Coefficient (/mm)
0.03 1 mm
2 mm
3 mm
0.02
0.01
0.00
0.01
0.02
0 10 20 30 40 50
x (mm)
FIGURE 2.5 Recovered optical absorption profiles plotted along y = 1 mm (a), y = –7

mm(3 mm-diameter target) and y = 8 mm(2 mm-diameter target) (b), and y = 1 mm (c) from
the images shown in Figures 2.4(a)–(c), respectively.
× 103
20
20
15
10
10
0
5
10
0
20
20 10 0 10 20
(a)
× 103
20
10
10
0 5
10
0
20
20 10 0 10 20
(b)
FIGURE 2.6 Reconstructed optical absorption coefficient images (units: mm–1). A 2 mm-
diameter target having an optical contrast of 2:1 (a) and 1.5:1 (b), relative to the background.
The reconstructed absorption coefficient images of the final two cases are
shown in Figures 2.6a and 2.6b, and the associated absorption coefficient profiles
are depicted in Figure 2.7. We immediately see that the different optical contrast
levels of the objects relative to the background are clearly resolvable.
We also observe that the imaging quality for the smallest target (Figure 2.4c) and
lowest contrast (Figure 2.6b) cases is degraded with stronger artifacts and over- or
underestimated target size, compared with that for the larger target and higher
contrast cases. The degradation is most likely due to the lower signal-to-noise ratio
(SNR) for the smallest target and lowest contrast cases. We also believe that the
nonuniformity, significant variation, and negative values seen in Figures 2.4 and
2.6 are primarily caused by the limited bandwidth of the transducer used, which
is directly related to the target size. This is evident from Figure 2.5 where we see
a clear trend—the smaller the target size, the larger the amplitude of the negative
values. For example, the peak negative value for the 0.5, 1.0, 2.0, and 3.0 mm target is
–0.02, –0.015, –0.007, and 0.003 mm –1, respectively. In addition, we found that
the negative value issue does not appear to be related to the contrast level. For
Absorption Coefficient(/mm)
0.020 1.5x
2.0x
0.016
0.012
0.008
0.004
0.000
0.004
0 10 20 30 40 50
x (mm)
FIGURE 2.7 Recovered optical absorption profiles plotted along y = 6 mm (a), and y =
6.5 mm (b) from the images shown in Figures 2.6(a) and (b), respectively.
example, the image of the 2 mm target size with 3:1 contrast (in Figure 2.5) gives
almost the same level of amplitude of negative value as the images of the 2 mm
target size with 2:1 or 1.5:1 contrast (Figure 2.7).
The assumption of homogeneous or constant absorption coefficient during the
procedure for estimating the distribution of optical fluence should not contribute
to the degradation significantly, because the heterogeneity size is small in the
cases studied here. We expect, however, that this assumption may have significant
impact on the estimation of the distribution of optical fluence if the heterogeneity
size becomes large (e.g., larger than 1 cm in diameter). In this case it would be
necessary to develop new PAT methods without such an assumption for recon-
structing the distribution of absorption coefficient (see Methods 3 and 4 below).

While Methods 1 and 2 described above are very effective in recovering optical
absorption coefficient, there are certain limitations associated with these approaches.
For example, for Method 1, one has to know the exact boundary reflection coefficients
as well as the exact strength and distribution of incident light source, which require
careful experimental calibration procedures. It is often difficult to obtain these initial
parameters accurately. The use of Method 2 requires diffusing light measurements,
making the hardware implementation complicated. In addition, the recovered results
for both Methods 1 and 2 strongly depend on the accuracy of the distribution of abso-
lute absorbed energy density from conventional PAT. Method 3 is able to overcome
these limitations by combining conventional PAT with the photon diffusion equation-
based regularized Newton method for accurate recovery of optical absorption coeffi-
cient. This approach is based on the following photon diffusion equation as well as the
Robin boundary conditions, in consideration of the absorbed energy density, Ψ = a Φ,
⋅ D(r ) ( E (r )Ψ(r )) − Ψ(r ) = − S (r ) (2.5)
− D ( E (r )Ψ) ⋅ n = E (r )αΨ (2.6)

where E (r ) = 1/ a (r ), and S(r) is the incident point or distributed source term. For the
inverse computation, the so-called Tikhonov regularization sets up a weighted term
as well as a penalty term to minimize the squared differences between computed and
measured absorbed energy density values,
2 2
min Ψ c − Ψ o + β L[ E − E0 ] (2.7)
{
χ
where L is the regularization matrix or filter matrix, and β is the regularization

parameter. Ψ o = ( Ψ 1o , Ψ o2 ,..., Ψ oN )T and Ψ c = ( Ψ 1c , Ψ c2 ,..., Ψ cN )T , where Ψ io is the
absorbed energy density obtained from PAT, and Ψ ic is the absorbed energy density
computed from Equations (2.5) and (2.6) for i = 1, 2…, N locations within the entire
PAT reconstruction domain. The initial estimate of absorption coefficient can be
updated based on iterative Newton method as follows,
(δE) = ( JT J + λI + βLT L)−1 [ JT ( Ψ o − Ψ c ) − βLT LE] (2.8)
where J is the Jacobian matrix formed by ∂Ψ/∂E inside the whole reconstruction
domain including the boundary zone. The practical update equation resulting from
Equation (2.8) is utilized with β = 1,
(E) = ( JT J + λI + LT L)−1 [ JT ( Ψ o − Ψ c )] (2.9)
In addition to the usual Tikhonov regularization, the PAT image (absorbed energy
density map) is used both as input data and as a priori structural information to
regularize the solution so that the ill-posedness associated with such inversion can
be reduced. In our reconstruction scheme, the PAT image is first segmented into dif-
ferent regions according to different heterogeneities or tissue types. We then employ
both the distribution of absorbed energy density in the entire imaging domain and
segmented a priori structural information for optical inversion. The segmented
a priori spatial information can be incorporated into the iterative process using the
regularization filter matrix, L shown in Equation (2.9). Laplacian-type filter matrix is
employed and constructed according to the region or tissue type it is associated based
on derived priors. This filter matrix is able to relax the smoothness constraints at the
interface between different regions or tissues, in directions normal to their common
boundary so that the covariance of nodes within a region is basically realized. As
such, the elements of matrix L, Lij, are specified as follows (Yuan et al. 2007):
1 if i= j
Lij = −1/ NN if i, j one region (2.10)
0 if i, j diffrent region

where NN is the total node number within one region or tissue. It is noted from
Equation (2.9) that a hybrid regularization scheme that combines both Levenberg–
Marquardt and Tikhonov regularizations are utilized. The optical absorption coef-
ficient distribution is reconstructed through the iterative procedures described by
Equations (2.5), (2.6), and (2.9).
Examples
Experimental data used to validate Method 3 were collected using the PAT system
presented in Figure 1.1. For the first two experiments, we embedded two objects
with a size ranging from 2.0 to 5.5 mm in diameter in a 50.8 or 40.0 mm-diameter
solid cylindrical phantom. We then immersed the object-bearing solid phantom
into a 110.6 mm-diameter water background. The phantom materials used con-
sisted of Intralipid as scatterer and India ink as absorber with Agar powder (1–2%)
for solidifying the Intralipid and India ink solution. The background phantom had
μa = 0.01 mm–1 and μs′= 1.0 mm–1 while the two targets had μa= 0.03 mm–1 and μs′ =
2.0 mm–1 for test 1, and μa= 0.07 mm–1 and μs′ = 3.0 mm–1 for test 2. In the next two
experiments, we placed a single-target-containing phantom into the water, aiming
to test the capability of resolving target having different optical contrasts relative
to the background phantom. The target size was 1.0 and 2.0 mm in diameter for
tests 3 and 4, respectively. The target had μa= 0.03 mm–1 and μs′= 2.0 mm–1 for test
3, and μa= 0.015 mm–1 and μs′ = 2.0 mm–1 for test 4. In the image reconstructions
for the four experiments, we assumed scattering coefficient known as constant
(1.0mm–1). The initial guesses of optical absorption coefficient for the target(s) and
background medium were 0.02 mm–1 and 0.01 mm –1, respectively.
The results from the first two sets of experiments are shown in Figure 2.8, where
(a) and (b) present the reconstructed absorption coefficient images of two objects
having a size of 2.0 and 3.0 mm (test 1), and 5.5 mm (test 2) in diameter, respec-
tively, while the recovered absorbed energy density maps for experiments 1 and 2
are also plotted in (c) and (d) for comparison. We see that the objects in each case
are clearly detected. As shown in Table 2.1, the recovered absorption coefficients
20
15 0.025
10 0.02
5
0 0.015
5
10 0.01
15 0.005
20
25
20 10 0 10 20
(a)
FIGURE 2.8 Reconstructed absorption coefficient images (a, b), absorbed light energy den-
sity images (c, d). (a), (c) are for experiment 1, and (b), (d) for experiment 2. The axes (left
and bottom) illustrate the spatial scale, in mm, whereas the scale (right) records the absorbed
optical energy density in relative units, or absorption coefficient in mm–1. (continued)
20 0.07
0.06
10
0.05
0 0.04
0.03
10
0.02
20
20 10 0 10 20
(b)
20 6000
5000
10 4000
0 3000
2000
10
1000
20 0
20 10 0 10 20
(c)
20
0.025
10
0.02
0.015
0
0.01
10
0.005
20
20 10 0 10 20
(d)
FIGURE 2.8 (continued) Reconstructed absorption coefficient images (a, b), absorbed light
energy density images (c, d). (a), (c) are for experiment 1, and (b), (d) for experiment 2. The
axes (left and bottom) illustrate the spatial scale, in mm, whereas the scale (right) records the
absorbed optical energy density in relative units, or absorption coefficient in mm–1.
of the target and background are quantitative compared to the exact values for both
experiments. By estimating the FWHM of the absorption coefficient profiles, the
recovered object size was found to be 1.8, 2.7, and 5.0 mm, which is also in good
agreement with the actual object size of 2.0, 3.0, and 5.5 mm for experiments 1
and 2. The reconstructed absorption coefficient images for experiments 3 and 4
are shown in Figures 2.9a and 2.9b. We immediately see that the different optical
contrast levels of the objects relative to the background are quantitatively resolved.
TABLE 2.1
Average Value of the Recovered Absorption Coefficient (mm –1) of Target and
Background and Target Size (mm) for Experiments 1 and 2
Targets (μa) Background(μa) Target Sizes
Test Reconstructed Exact Reconstructed Exact Reconstructed Exact
(1) Target 1 0.027 0.03 0.012 0.01 1.8 2.0
Target 2 0.028 0.03 0.012 0.01 2.7 3.0
(2) Target 1 0.067 0.07 0.011 0.01 5.0 5.5
It is important to note that our reconstruction method does not need any
calibration procedure due to the use of relative incident laser source strength
and normalized absorbed energy density distribution where the normalization is
performed by simply dividing the absorbed energy density at each nodal location
with the maximum absorbed energy density (i.e., Ψ(r ) = Ψ(r )/Ψmax). An optimiza-
tion scheme was then applied to search for the boundary conditions coefficient,
20
0.025
10
0.02
0
0.015
10
0.01
20
20 10 0 10 20
(a)
× 103
20 14
10 12
0 10
10 8
20 6
20 10 0 10 20
(b)
FIGURE 2.9 Reconstructed absorption coefficient images for experiment 3 (a) and experi-
ment 4 (b). The axes (left and bottom) illustrate the spatial scale in mm, whereas the scale
(right) records the absorption coefficient in mm–1.
α and the relative source strength as described previously (Iftimia and Jiang 2000).
As such, the reconstruction of optical properties with our algorithms does not
depend on the absolute values of absorbed energy density and optical fluence as
well as the boundary parameter. For example, even though the values/scales of
the absorbed energy density for experiments 1 and 2 are very different as shown
in Figures 2.8c and 2.8d, the algorithm is still able to recover the absorption coef-
ficient images quantitatively in terms of the location, size, and absorption coef-
ficient value of the objects. In addition, our method is able to resolve the issue of
negative absorbed energy density values often seen in conventional PAT.

Method 4 represents the best among all the four methods described in this chapter—
it requires only one-step to directly recover optical absorption coefficient from pho-
toacoustic measurements along boundary domain. The photoacoustic wave equation
in frequency-domain for an acoustically homogeneous medium is rewritten here:
2 v0βΨ
p(r , ω ) + k02 p(r , ω ) = ik0 (2.11)
Cp

The finite element discretization of Equation (2.11) is stated as,
Ap = B (2.12)
Based on the finite element solution to Equation (2.12), we can use Marquardt–
Tikhonov regularization-based Newton method to directly recover/update absorption
coefficient from an initial guess of μa,0 by minimizing an object function composed
of a weighted sum of the squared difference between computed and measured acous-
tic pressure around the boundary area given by
Min: F = ∑( p o
i − pic )2 (2.13)
i =1
As a consequence, the following matrix equation capable of inverse solution of μa,

is obtained for the inverse calculation:
( ℑT ℑ + λI) χ = ℑT ( po − pc ), a = a ,0 + ζ χ (2.14)

where po = ( p1o , p2o ,..., pMo )T and pc = ( p1c , p2c ,..., pM

c T
) in which pio and pic are observed
and computed normalized acoustic measurements for i = 1, 2…, M boundary loca-
tions; Δχ is the update vector for μa; ℑ is the Jacobian matrix formed by ∂ p/∂ a at the
boundary measurement sites; λ is a scalar; ζ is calculated from common backtrack-
ing line search (Yuan and Jiang 2007); and I is the identity matrix. In consideration
of Ψ = a Φ, the sensitivity of pressure to absorption coefficient (∂ p/∂ a ) equals to

the product of the sensitivity of pressure to energy density (∂p/∂Ψ) and the sensitivity
of energy density to absorption coefficient (∂Ψ/∂ a ). Specifically, the elements in
Jacobian matrix ℑ are determined by
N
∂ pi ∂ pi ∂Ψ k
∂ a, j
= ∑ ∂Ψ k ∂ a , j
(i = 1,2... M ; j, k = 1,2... N ) (2.15)
k =1
in which N is the nodal number of the finite element mesh in the entire problem domain
and the sensitivities of ∂p/∂Ψ in each iteration can be calculated from Equation (2.12)
based on adjoint method (see Section 3.2 for detail about this method),
∂p ∂B ∂A
[ A] = − {P} (2.16)
∂Ψ ∂Ψ ∂Ψ
The derivatives ∂Ψ/∂ a in Equation (2.15) are further written as
∂Ψ k Φk + a,k (∂Φ k /∂ a, j ) if (k = j)
= (2.17)
∂ a, j a,k (∂Φ k /∂ a, j ) if ( k ≠ j)

And photon fluence Φ in Equation (2.17) can be computed by the finite element
solution to the following photon diffusion equation:
⋅ D(r ) Φ(r ) − (r )Φ(r ) = − S (r ) (2.18)

α
in which S(r) is the normalized light source term and D(r) is the diffusion coefficient.
The finite element discretized form of Equation (2.18) is written,
[ K ]{Φ} = {b} (2.19)
Based on Equation (2.19), the sensitivity ∂Φ/∂ a in Equation (2.17) is obtained

from the following equation:
[ K ]{∂Φ/∂ a } = {∂b/∂ a } − [∂ K /∂ a ]{Φ} (2.20)

Thus the absorption coefficient distribution can be reconstructed through a

Newton iterative solution procedure described by Equation (2.12) (forward solution)
and Equation (2.14) (inverse solution). The regularization parameter is determined
by combined Marquardt and Tikhonov regularization schemes. We have found that
when λ = P0 – Pc × trace[JTJ], the reconstruction algorithm generates best results for
PAT image reconstruction.
Examples
Phantom experiments were performed to validate this one-step quantitative PAT

method using the imaging system shown in Figure 1.1. In the experiments, we
embedded one or two objects with a size ranging from 0.5 to 3.0 mm in the 5 cm-
diameter solid cylindrical phantom (Intralipid + India ink + Agar powder). We then
immersed the object-bearing solid phantom into the water tank. The absorption of
the background phantom was 0.01 mm–1, while the absorption coefficient of the
target(s) was 0.03 mm–1. The reduced scattering coefficients of the background
medium and target(s) were 1.0 and 3.0 mm –1, respectively. As a final test, we
placed human hair-containing phantom (4 hairs) into the water tank to show the
high-resolution imaging capability of the developed scheme.
The results from the first experiment are shown in Figure 2.10, where (a) and
(b) show the recovered μa maps using Method 1 and Method 4, respectively. By
estimating the FWHM of the absorption coefficient profiles for images (a) and
(b), we found the recovered object sizes were 1.9 and 2.9 mm, respectively,
which are in good agreement with the actual object sizes of 2.0 and 3.0 mm.
Figure 2.11 plots the reconstructed absorption coefficient images for experimen-
tal test 2 using both Method 1 and Method 4. The recovered target size using
Method 4 and Method 1 was found to be 0.55 and 0.8 mm, respectively, com-
pared to the actual target size of 0.5 mm. Obviously Method 1 overestimated the
target size in most cases. We also found from Figures 2.10 and 2.11 that Method
4 is able to considerably reduce the effect of local optical fluence and boundary
noise. Further, it is observed from Figure 2.12 that Method 4 can clearly image
human hairs embedded in phantom with submillimeter resolutions and quanti-
tatively accurate absorption coefficient.
We have demonstrated experimental evidence that it is possible to directly
recover the absolute μa image using boundary photoacoustic measurements cou-
pled with the photon diffusion equation in just one step. The method described
is able to quantitatively reconstruct absorbing objects with different sizes and
optical contrast levels. In particular, it is noted that this method does not need
any calibration procedure because it allows for the use of relative incident laser
source strength and normalized boundary measurements of acoustic pressure.
The source intensity and boundary condition coefficient can be optimized and
localized based on the photon diffusion equation by minimizing f with the normal-
ized acoustic measurements p : f = ∑ M 2
i =1(( aΦ)i − pi ) . As such, the reconstruction
of absorption coefficient with Method 4 does not depend on the absolute values
of absorbed energy density and optical fluence, whereas Methods 1–3 are all
dependent on these values. It should be pointed out, however, that the choice of
the initial guess of μa does have some effect on the recovered results. To resolve
this issue, we have used an optimization scheme adapted from a previous study of
diffuse optical tomography (Iftimia and Jiang 2000) to search for an optimal initial
guess of μa through a forward fitting procedure based on the time-domain photo-
acoustic wave equation. In addition, we note that different transducer sensitivity
may lead to a different scaling factor of the reconstructed absorption coefficient
images. Routine calibrations on the transducer response need to be done when
using Methods 1–3. However, in Method 4 we use normalized acoustic data in
arbitrary units (i.e., the largest value of pressure is 1.0) to recover the absorption
coefficient, making the impact of transducer sensitivity minimal. This impact is
further implicitly minimized through the iteration procedure of the method.
20 20
10 10
0 0
–10 –10
–20 –20
–20 0 20 –20 –10 0 10 20

(a) (b)
0.01 0.02 0.03
0.035
New scheme
0.030 Previous scheme
0.025
0.020
0.015
0.010
0.005
0.000
–0.005
–20 –10 0 10 20
x (mm)
(c)
FIGURE 2.10 Reconstructed μa image using Method 1 (a), Method 4 (b), and recovered
optical absorption profile (c) plotted along y = –7.0 mm from the images shown in Figures
2.10(a) and (b). The axes (left and bottom) illustrate the spatial scale, in mm, whereas the
color scale (right) records μa in mm–1.
20
20
10
10
0 0
–10 –10
–20 –20
–20 0 20 –20 –10 0 10 20

(a) (b)
0.01 0.02 0.03
0.035
0.030 New scheme
Previous scheme
0.025
0.020
0.015
0.010
0.005
0.000
–0.005
–0.010
–0.015
–0.020
–0.025
–20 –10 0 10 20
x (mm)
(c)
FIGURE 2.11 Reconstructed μa images using Method 1 (a), Method 4 (b), and recovered opti-
cal absorption profile (c) plotted along y = 1.0 mm from the images in Figures 2.11(a) and (b).
10
8
0.25
6
4
0.2
2
0 0.15
2
0.1
4
6 0.05
8
10
10 5 0 5 10
FIGURE 2.12 The recovered μa image for the human hairs embedded in phantom.
2.5 SIMULTANEOUS RECOVERY OF OPTICAL ABSORPTION

COEFFICIENT AND ACOUSTIC VELOCITY
As we indicated previously, the recovery of an optical absorption coefficient allows
us to obtain physiologically relevant tissue parameters including deoxy- and oxy-
hemoglobin and water content. Here we show we can reconstruct both acoustic
velocity and absorbed light energy density based on the PA wave equation for an
inhomogeneous acoustic medium (see Equation 1.23). This ability of recovering both
optical and acoustic properties not only provides more accurate reconstruction of
optical property over the qualitative PAT because of the elimination of the assump-
tion of homogenous acoustic velocity built in the qualitative PAT methods, but also
adds the potential to better differentiate benign from malignant lesions as it is known
that significant differences exist in acoustic properties between normal and tumor
tissues (Greenleaf et al. 1981).
The frequency-domain PA wave equation for an acoustically heterogeneous
medium is rewritten here,
2 v0βΨ(r )
p(r , ω ) + k02 (1 + O) p(r , ω ) = ik0 (2.21)
Cp

where O is a coefficient that depends on both acoustic speed and attenuation as fol-
lows (Jiang et al. 2006):
co2 i αv0
O= 2
−1+ (2.22)
c k0 v 2
in which v is the speed of acoustic wave in the medium/tissue and α is the acoustic
attenuation coefficient. The finite element discretization of Equation (2.21) is stated as,
Ap = B (2.23)
where the elements of the matrix A and B are expressed as
Aij = ψ j ⋅ ψ i − k02 ψ j ψ i − k02 ∑O

k
R,k ψ k ψ jψi
∂2 ψ j
− ik02 ∑ l
OI ,l ψ l ψ j ψ i −
∫ ηψ j + γ
∂φ2
ψ i ds,

v0β v0β
Bi = −ik0
Cp ∑Ψ
k
R ,k ψ k ψ i + k0
Cp ∑Ψ
l
I ,l ψ l ψ i (2.24)

in which Ψ, O, and p are expressed by their real and imaginary components, Ψi is the
basis function, < > indicates integration over the problem domain; and the second-
order absorption boundary conditions are employed here (Yuan et al. 2005),
∂2 ψ n
ψ n ⋅ nˆ = ηψ n + γ (2.25)
∂2 θ
2
− ik − 3/2 ρ+ i 3/8 kρ− i/2 kρ2
where η = 1− i/kρ , γ = 1− i/kρ and θ is the angular coordinate at a radial posi-
tion ρ. To obtain a matrix equation capable of inverse solution, a combined Marquardt
and Tikhonov regularization scheme is used for the inverse calculation,
(ℑT ℑ + λI) χ = ℑT ( po − pc ) (2.26)

where po = ( p1o , p2o ,..., pMo )T , pc = ( p1c , p2c ,..., pM ) , and pio and pic are observed
c T
and computed complex acoustic field data for i = 1, 2…, M boundary measurement
locations, Δχ is the update vector for the optical and acoustic properties, ℑ is the
Jacobian matrix formed at the boundary measurement sites, λ is a scalar, and I is the
identity matrix. Thus here the image formation task is to update optical and acoustic
property distributions via iterative solution of Equations (2.23) and (2.26) so that a
weighted sum of the squared difference between computed and measured acoustic
pressure can be minimized.
Examples
Phantom experiments were conducted using the PAT system shown in Figure 1.1
to demonstrate this reconstruction approach. In the first two experiments, we
embedded one glycerol powder-containing target (10 mm in diameter) in a 30 mm-
diameter solid cylindrical phantom. We then immersed the object-bearing solid
phantom in a 50.8 mm-diameter water background, as displayed in Figure 2.13a.
The acoustic speed in the region of the target containing 10% (phantom 1) and
Φ = 10 2 Φ = 10
3.5 X
Φ=5 2 7
r = 1.5
Transducer
path
(a) (b) (c)
FIGURE 2.13 Experiment geometry for (a) phantoms 1 and 2, (b) phantom 3, and (c) phan-
tom 4. The dimension units are millimeters.
28% (phantom 2) glycerol was 1540 and 1620 m/s, respectively, while the acoustic
speed for the background was 1480 m/s. For the third experiment, a 5 mm- and
a 10 mm-diameter target (each containing 28% glycerol) were embedded into a
30 mm-diameter solid cylindrical phantom. Again the entire solid phantom was
immersed in a 50.8 mm-diameter water background, as shown in Figure 2.13b.
In the final experiment, we embedded two nanoparticle-containing objects
(3 mm in diameter) in a 50 mm-diameter solid cylindrical phantom, as plotted
in Figure 2.13c. Each target had 0.5 nM nanoparticles. We then immersed the
object-bearing solid phantom in a 110.6 mm-diameter water background. For all
the experiments, the phantom materials used consisted of Intralipid as scatterer
and India ink as absorber with Agar powder for solidifying the Intralipid and India
ink solution. The optical absorption and scattering coefficients for the phantom
were 0.004 and 1.0 mm –1, respectively. The optical absorption contrast between
the target and phantom is about 10 for experiments 1–3, and 20 for experiment 4.
Figures 2.14a and 2.14b show the reconstructed acoustic and optical images for
experiment 1, while 2.14c and 2.14d present the recovered acoustic and optical
× 106
20 1.525
1.52
10
1.515
0 1.51
10 1.505
1.5
20 1.495
20 10 0 10 20
(a)
0.16
20
0.14
10 0.12
0 0.1
0.08
10
0.06
20
0.04
20 10 0 10 20
(b)
FIGURE 2.14 Reconstructed acoustic speed (a, c) and absorbed optical energy density
(b, d) images with different acoustic contrast levels with respect to the background, (a) and
(b) for experiment 1, and (c) and (d) for experiment 2. The axes (left and bottom) illustrate the
spatial scale in millimeters, whereas the scale (right) records the acoustic speed, in mm/s, or
absorbed optical energy density, in a relative unit. (continued)
× 106
1.65
20
1.6
10
1.55
0
1.5
10
1.45
20
1.4
20 10 0 10 20
(c)
20 0.25
10 0.2
0.15
0
0.1
10
0.05
20
0
20 10 0 10 20
(d)
FIGURE 2.14 (continued) Reconstructed acoustic speed (a, c) and absorbed optical energy
density (b, d) images with different acoustic contrast levels with respect to the background,
(a) and (b) for experiment 1, and (c) and (d) for experiment 2. The axes (left and bottom) illus-
trate the spatial scale in millimeters, whereas the scale (right) records the acoustic speed, in
mm/s, or absorbed optical energy density, in a relative unit.
images for experiment 2. We see that the object in each case is clearly detected.
By estimating FWHM of the acoustic property, as displayed in Figure 2.15, we
found that the recovered size of these objects ranged from 9.1 to 9.3 mm, in good
agreement with the actual object size of 10 mm. In our experiments, the 10% and
28% glycerol-containing target used gave an acoustic speed of 1540 and 1620 m/s,
respectively. As can be seen from Figure 2.15, we are able to obtain quantitatively
resolved images in terms of the acoustic property as well as the position, shape,
and size of the objects.
Figure 2.16a and 2.16b displays the reconstructed acoustic speed and optical
images for the third experiment having two targets. By estimating the FWHM of
the acoustic profiles plotted in Figure 2.17, we found that the recovered size of the
two objects was 4.6 and 9.8 mm, respectively. Again, both size values were in
good agreement with the exact values of 5 and 10 mm. As shown in Figure 2.16,
the reconstructed acoustic properties for these cases are also accurate relative to
1,530,000
1,525,000
1,520,000
Velocity (mm/s) 1,515,000
1,510,000
1,505,000
1,500,000
1,495,000
1,490,000
15 10 5 0 5 10 15
y (mm)
(a)
1,640,000
1,620,000
1,600,000
Velocity (mm/s)
1,580,000
1,560,000
1,540,000
1,520,000
1,500,000
1,480,000
15 10 5 0 5 10 15
x (mm)
(b)
FIGURE 2.15 Recovered acoustic speed profiles plotted along x = 0 mm (a), y = 0 mm

(b) from the images shown in Figures 2.14(a) and (c), respectively.
the exact value, which is further confirmed by the quantitative analysis shown in
Figure 2.17.
Figures 2.18a and 2.18b present the reconstructed acoustic and optical prop-
erties of two objects having 0.5 nM nanoparticles within each of the objects. By
estimating the FWHM of the acoustic speed plotted in Figure 2.18c, we found that
the recovered diameter size of these objects ranges from 2.8 to 3.3 mm, in good
agreement with the actual object size of 3 mm. In addition, it is noted the imaging
resolution in Figure 2.16 is not as good as that in Figure 2.18 due to the low opti-
cal contrast between the targets and phantoms for experiments 1–3. Further, from
Figures 2.14, 2.16, and 2.18 we can also see that the image quality of both opti-
cal and acoustic images appears to be similar when different acoustic contrasts
between the target and the background were used.
× 106
1.62
20
1.6
10 1.58
1.56
0
1.54
10 1.52
20 1.5
1.48
20 10 0 10 20
(a)
20 0.2
10 0.15
0 0.1
10
0.05
20
0
20 10 0 10 20
(b)
FIGURE 2.16 Reconstructed acoustic speed (a) and absorbed optical energy density
(b) images for experiment 3. The axes (left and bottom) illustrate the spatial scale in millime-
ters, whereas the scale (right) records the acoustic speed, in mm/s, or absorbed optical energy
density, in a relative units.
1,650,000
1,600,000
1,550,000
Velocity (mm/s)
1,500,000
1,450,000
1,400,000
1,350,000
1,300,000
15 10 5 0 5 10 15
x (mm)
(a)
FIGURE 2.17 Recovered acoustic speed profiles plotted along y = 0 mm (a), y = –8 mm

(b) from the image shown in Figure 2.16(a). (continued)
1,650,000
1,600,000
Velocity (mm/s)
1,550,000
1,500,000
1,450,000
1,400,000
–15 –10 –5 0 5 10 15
x (mm)
(b)
FIGURE 2.17 (continued) Recovered acoustic speed profiles plotted along y = 0 mm (a), y
= –8 mm (b) from the image shown in Figure 2.16(a).
× 106
50
1.6
1.55
0 1.5
1.45
50 1.4
50 0 50
(a)
0.35
50
0.3
0.25
0.2
0
1.15
0.1
0.05
50
50 0 50
(b)
FIGURE 2.18 Reconstructed acoustic speed (a) and absorbed optical energy density
(b) images for experiment 4. (c) is the recovered acoustic speed profile plotted along x = 0 mm
from the image shown in (a).The axes (left and bottom) illustrate the spatial scale in millime-
ters, whereas the scale (right) records the acoustic speed, in mm/s, or absorbed optical energy
density, in a relative units. (continued)
1,620,000
1,600,000
1,580,000
1,560,000
Velocity (mm/s)
1,540,000
1,520,000
1,500,000
1,480,000
1,460,000
1,440,000
15 10 5 0 5 10 15
y (mm)
(c)
FIGURE 2.18 (continued) Reconstructed acoustic speed (a) and absorbed optical energy
density (b) images for experiment 4. (c) is the recovered acoustic speed profile plotted along
x = 0 mm from the image shown in (a).The axes (left and bottom) illustrate the spatial scale in
millimeters, whereas the scale (right) records the acoustic speed, in mm/s, or absorbed optical
energy density, in a relative units.
2.6 MULTISPECTRAL PAT
For most of the cases where the heterogeneity/object has contrast in both the optical
property and acoustic velocity, the absorbed energy density and acoustic velocity can
be recovered at a single wavelength. This is very similar to that seen in diffuse optical
tomography (DOT) where both absorption and scattering coefficients can be recon-
structed at a single wavelength when the objects have contrast in both the absorption
and scattering (Jiang 2010). For such cases, crosstalk certainly exists, but it is not strong.
This crosstalk is also dependent on the level of contrast for the two parameters. For the
experimental cases examined in Sections 2.1–2.5, the level of contrast used was rela-
tively low, which explains how the insignificant crosstalk occurred in our reconstruction.
However, for the cases where acoustic heterogeneity is quite different from the
distribution of absorbed light energy density, our simulation tests indicate that the
crosstalk cannot be minimized using one wavelength measurement. For our simula-
tion tests, a circular background region (15 mm in radius) contains three circular
targets (2 mm in radius each) positioned at 2 (top left), 9 (top right), and 6 o’clock
(bottom). Top left inclusion has contrast for velocity, while bottom and top right
inclusions have contrast for absorbed energy density. The velocity for the target
and background is 1.54 × 106 mm/s and 1.44 × 106 mm/s, respectively, while the
absorbed energy density for the background and targets is 1 mJ/mm3 and 10 mJ/mm3,
respectively. We can see from the recovered images shown in Figure 2.19 that the
absorbed energy density distribution has some crosstalk effect on acoustic veloc-
ity, while acoustic velocity distribution has no significant effect on the recovered
absorbed energy density due to its low contrast.
15
10
y 0
–5
–10
–15
–10 0 10
x
(a)
× 106
15
1.5
10 1.48
5 1.46
1.44
0
1.42
5
1.4
10 1.38
15 1.36
15 10 5 0 5 10 15
(b)
15
10
y 0
–5
–10
–15
–10 0 10
x
(c)
FIGURE 2.19 Reconstructed images using one wavelength measurement. (a) and (b) are
the exact and recovered acoustic velocity (mm.s–1). (c) and (d) are the exact and recovered
absorbed energy density(mJ.mm–3). (continued)
15
10
10
5 8
0 6
5 4
10
2
15
10 0 10
(d)
FIGURE 2.19 (continued) Reconstructed images using one wavelength measurement.

(a) and (b) are the exact and recovered acoustic velocity (mm.s–1). (c) and (d) are the exact and
recovered absorbed energy density(mJ.mm–3).
Thus, in most cases, multispectral photoacoustic measurements are required

when both optical and acoustic properties are to be recovered. In multispectral
PAT, frequency-domain Helmholtz wave equation in an acoustically heterogeneous
medium is written as
2 v 0β a (r , λ)Φ(r , λ)
p(r , ω , λ) + k02 (1 + O) p(r , ω , λ) = ik0 (2.27)
Cp

where λ is the wavelength of incident light. In addition, according to the Beer’s law,
the wavelength-dependent tissue absorption can be expressed as
a (λ ) = ∑ ε (λ)c (2.28)
i i
i =1
where ci is the concentration and εi (λ) is the extinction coefficient of the ith chro-
mophore (HbO2, Hb or H2O) at wavelength λ. In consideration of Equation (2.28),
Equation (2.27) can be rewritten,
2 2
v0 β ∑ ε (λ)c Φ(r , λ)
i =1
i i
p(r , ω , λ) + k (1 + O) p(r , ω , λ) = ik0

0 (2.29)
Cp

Thus, similar to Equations (2.14) and (2.26), the inverse solution can be obtained
by solving the following equation:
( J T J + λ I) χ = J T ( po − pc ) (2.30)

in which χ = ( O1 , O2 ,…, On , c1,1 , c1,2 ,…, c1,n , c2,1 , c2,2 ,…, c2,n , c3,1 ,
c3,2 ,…, c3,n )T is the update vector for chromophores and acoustic velocity; λ′ is
the regularization parameter determined by combined Marquardt and Tikhonov
regularization schemes; pio and pci are measured and calculated data for i = 1,2,…M
boundary locations and are written as for each acoustic frequency ω and each optical
wavelength λ,
po = ( p1o , p2o ,…, pMo )T (ω , λ), pc = ( p1c , p2c ,…, pM ) (ω , λ) (2.31)

c T

The Jacobian matrix, J, is denoted as J = [ J O ,λ ,ω , J ci ,λ ,ω ], where J O ,λ ,ω and J ci ,λ ,ω

represent the Jacobian submatrix for acoustic velocity and different chromophores,
respectively. It should be noted that in this reconstruction algorithm we assume the
optical fluence Φ can be estimated through the photon diffusion equation during
image reconstruction procedure (see Section 2.1).
Examples
Here we use numerical simulations to demonstrate the feasibility of the multispec-

tral PAT described above. For the simulations, a circular background (15 mm in
radius) contained three circular targets (2mm in radius each), where each inclusion
had different contrast for each given parameter including Hb, HbO2, H2O and
acoustic velocity. The chromophore concentrations and velocity used for all the
test cases are listed in Table 2.2. Six optical wavelengths (633, 682, 723, 850, 854,
and 930 nm) were chosen for the spectral resolved PAT. The “measured” data were
generated using a forward computation with 2% random Gaussian noise. A total
of 120 receivers were equally distributed along the boundary of the circular back-
ground, and 50 frequencies ranging from 50 to 540 kHz were used at each optical
wavelength for each receiver. The extinction coefficient for each chromophore
used was taken from Pahl (2003). The images in the left column of Figure 2.20 show
TABLE 2.2
Chromophore Concentrations and Acoustic Velocity for Test Objects
Object Position HBO2 (μM) HB (μM) H2O(%) Acoustic Velocity (mm/s)
Background 6 2 20 1.44×106
Top right 20 12 50 1.44×106
Top left 6 12 20 1.44×106
Bottom middle 20 12 50 1.74×106
15 15
12
10 10
10
5 5 8
y 6
0 0
–5 –5 4
–10 2
–10
0
–15 –15
–10 0 10 –10 0 10
x
15 15
10 10
20
5 5
y 15
0 0
–5 –5 10
–10 –10
5
–15 –15
–10 0 10 –10 0 10
x
15 15
10 50
10
5 45
5
40
y
0 0 35
–5 –5 30
–10 –10 25
20
–15 –15
–10 0 10 –10 0 10
x
15 × 106
15
1.7
10 10
5 5 1.65
y 0 0 1.6
–5 –5
1.55
–10 –10
1.5
–15 –15
–10 0 10 –15 –10 –5 0 5 10 15
x
FIGURE 2.20 Exact (left column) and reconstructed (right column) images using six opti-
cal wavelengths. The first to the fourth rows show the Hb, HBO2, H2O, and acoustic velocity
images, respectively.
the true locations of the targets including HbO2, HbR, H2O and acoustic velocity.
The reconstructed images shown in the right column of Figure 2.20 correspond to
HbO2, HbR, H2O and acoustic velocity, respectively. We see from Figure 2.20 that
the crosstalk errors between the chromophore concentrations and acoustic veloc-
ity have been effectively reduced using our quantitative PAT approach.
3 Image Enhancement
Software and Hardware
Approaches
In this chapter we describe a series of software approaches including dual meshing,

adjoint sensitivity, total variation minimization, radiative transfer equation, and par-
allel computation to enhance photoacoustic tomography (PAT) performance. These
software approaches aim to improve image reconstruction in terms of forward and
inverse solution accuracy, reduced noise effect, and computation speed. We also
present several hardware approaches to improve the signal-to-noise ratio (SNR) of
ultrasound transducer for improved PAT performance.
3.1 DUAL MESH SCHEME

Realizing the fact that acoustic fields at MHz frequencies change rapidly while
tissue optical and acoustic property distributions are usually relatively uniform,
a dual meshing method is required for the iterative nonlinear finite element (FE)-
based algorithms for fast yet accurate inverse computation. The dual meshing
scheme exploits two separate meshes—one fine mesh for accurate wave propa-
gation and one coarse mesh for parameter recovery. This dual meshing scheme
allows a significant reduction of the problem size during the reconstruction, thus
increasing overall computational efficiency. In fact, the idea of dual meshing has
been implemented in optical image reconstruction, where this method has proved
to be able to significantly enhance the quality of image reconstruction (Jiang et al.
1997, 1998; Gu et al. 2003). Here we present the implementation of this method in
PAT reconstruction.
Implementation of the dual meshing scheme impacts two components of the
reconstruction algorithm: (1) the forward solution at each iteration for the scatter-
ing pressure field where the acoustic and optical property profiles are defined on
the coarse mesh while the forward solution calculation is based on the fine mesh,
and (2) calculation of the Jacobian matrix, which is used to update the acoustic and
optical property profile estimates during the inverse solution procedure. Thus, for
the forward solution, the inner product, 〈(⋅)〉, e.g., in Equations (2.23) and (2.24), is
performed over the elements of the fine mesh, while OR, OI, ΨR, and ΨI need to be
expanded in the basis functions that are defined over the coarse mesh. For example,
47
for an arbitrary node i of the fine mesh which is embedded in a coarse mesh element
with nodes L1, L2, and L3, the values of OR and OI at node i become
OR ( xi , yi ) = ∑O
n =1
R , Ln φ Ln ( x , y)
OI ( xi , yi ) = ∑O I , Ln φ Ln ( x , y) (3.1)
n =1
where φ Ln is the Lagrangian basis function over the coarse mesh.

The second impact of the dual meshing method appears during the construction
of Jacobian matrix, J, which is used to update the object profile values. The elements
of J are composed of the partial derivatives of the scattering field at the observation
sites with respect to the values of OR, OI , ΨR, and ΨI at each node within the coarse
mesh. Considering the impact of the dual meshing, the elements of the Jacobian
matrix can be written as follows:
∂ Aij /∂OR ,k = − k02 φ k ψ i ψ j , ∂ Aij /∂OI ,l = −ik02 φl ψ i ψ j ,
∂ Aij /∂Ψ R ,k = −ik0 c0βφ k ψ i /C p , ∂ Aij /∂Ψ I ,l = k0 c0βφl ψ i /C p , (3.2)

where k and ℓ are the nodes on the coarse mesh, ϕk and φ are the basis functions
centered on nodes k and ℓ in this mesh, and the inner products are still performed
over the elements in the fine mesh. Since ϕk and ψ i are defined on the coarse mesh
and fine mesh, respectively, evaluating these inner products can be quite involved. A
way to simplify these integrations is to generate the fine mesh from the coarse mesh
by splitting the coarse elements into fine elements.
Examples
Two phantom experiments were conducted to test the dual meshing scheme. We
employed single-target-containing phantom tests, aiming to validate the accu-
racy improvement in detecting target when the dual mesh is used. The phantom
materials used consisted of Intralipid as scatter and India ink as absorber with
Agar powder (1–2%) for solidifying the Intralipid and India ink solution. The
absorption coefficient of the background phantom was 0.01 mm –1, while the
absorption coefficient of the target was 0.015 mm –1 and 0.02 mm –1 for test 1 and
test 2, respectively. Figure 3.1 shows the reconstructed absorbed energy density
images for the two cases with uniform mesh and dual mesh. We see that the tar-
get is clearly better recovered with the dual mesh (b and d) in terms of its shape
and size over that with the uniform mesh (a and c) for both cases. We also note
that the background region for both cases is more smoothly recovered using the
dual mesh scheme compared to that using the uniform mesh.
Image Enhancement Software and Hardware Approaches 49
14000
20 10000 10
12000
8000
10 5 10000
6000
8000
0 4000 0
6000
–10 2000 –5 4000
0
2000
–20 –10
–2000 0
–20 –10 0 10 20 –10 –5 0 5 10
(a) (b)
× 104
2
20 10
15000
10 5 1.5
10000
0 0 1
5000
–10 –5
0.5
0
–20 –10
0
–20 –10 0 10 20 –10 –5 0 5 10
(c) (d)
FIGURE 3.1 Reconstructed photoacoustic images. (a) and (b): The image for case 1 with
uniform and dual mesh. (c) and (d): The image for case 2 with uniform and dual mesh.
3.2 ADJOINT SENSITIVITY METHOD

The coupled complex adjoint sensitivity method can be utilized to efficiently deter-
mine the Jacobian matrix (e.g., see Equation (2.26)). Direct differentiation of both
sides of Equation (2.23) with respect to OR and ΨR gives, respectively,
[ A]{∂ P/∂OR } = −[∂ A/∂OR ]{P} (3.3a)

[ A]{∂ P/∂Ψ R } = {∂ B/∂Ψ R } − [∂ A/∂Ψ R ]{P} (3.3b)

An equivalent set of equations can be obtained for differentiation with respect to
OI and ΨI by replacing OR with O I in (3.3a) and ΨR with ΨI in (3.3b). The Jacobian
matrix can be calculated through the following steps:
First, we define a N × M matrix Ξ, and let Ξ satisfy the following relationship:
T
[ A] [Ξ] = [ d ] (3.4)
where the vector Δd has the unit value at the measurement sites/nodes and zero at
other nodes. Then we left multiply (3.3a) and (3.3b) with the transpose of [Ξ],

[Ξ ]T [ A]{∂ P/∂OR} = − [Ξ ]T [∂ A/∂OR ]{P} (3.5a)

[Ξ ]T [ A]{∂ P/∂Ψ R} = [Ξ ]T {∂ B/∂Ψ R} − [Ξ ]T [∂ A/∂Ψ R ]{P} (3.5b)
Equations (3.5a) and (3.5b) can be further written,
{∂ P/∂OR }T [ A]T [Ξ] = −{P}T [∂ A/∂OR ]T [Ξ] (3.6a)

{∂ P/∂Ψ R }T [ A]T [Ξ] = {∂ B/∂Ψ R }T [Ξ] − {P}T [∂ A/∂Ψ R ]T [Ξ] (3.6b)

Inserting Equation (3.4) into Equations (3.6a) and (3.6b), we get
{∂ P/∂OR }T = −{P}T [∂ A/∂OR ]T [Ξ] (3.7a)

{∂ P/∂Ψ R }T = {∂ B/∂Ψ R }T [Ξ] − {P}T [∂ A/∂Ψ R ]T [Ξ] (3.7b)

Now we can immediately tell that the left-hand side of the above equations actu-
ally gives the corresponding elements in the relative Jacobian matrix based on the
adjoint sensitivity method,
{∂ P/∂OR } = −[Ξ]T [∂ A/∂OR ]{P} (3.8a)

{∂ P/∂Ψ R } = [Ξ]T {∂ B/∂Ψ R } − [Ξ]T [∂ A/∂Ψ R ]{P} (3.8b)

3.3 TOTAL-VARIATION-MINIMIZATION SCHEME
Measurement noises are always the major factor affecting the quantitative accuracy
of the reconstructed images. For example, errors for quantitatively recovering opti-
cal absorption coefficient can be as large as 10% for simulated data with 5% noise
added and 20% for experimental data (Jiang et al. 2006; Yao and Jiang 2009). In
an effort to reduce the noise effect and to further enhance the quantitative accuracy
of photoacoustic image reconstruction, here we consider a total-variation-minimi-
zation scheme within our FE-based reconstruction framework. Our existing recon-
struction algorithms are based on the least squares criteria (i.e., the regularized
Newton method) (see Chapters 1 and 2) that stand on the statistical argument that
the least squares estimation is the best estimator over an entire ensemble of all
possible pictures. Total variation, on the other hand, measures the oscillations of
a given function and does not unduly punish discontinuities (Dobson and Santosa
1994; van den Berg and Kleinmann 1995). Hence, one can hypothesize that a
hybrid of these two minimization schemes should be able to provide higher quality
image reconstructions. In fact, the strategy of finding minimal total variation has
proved to be successful in applications including electrical-impedance tomography
(van den Berg and Kleinmann 1995), microwave imaging (Vogel and Oman 1996),
image processing (Dobson and Santosa 1994), optimal design (Acar and Vogel
1994), and diffuse optical tomography (Paulsen and Jiang 1996).
A practical need exists for reconstruction of photoacoustic images from few mea-
surements, as this can greatly reduce the required scanning time and the number of
ultrasound sensors placed near or on the boundary of an object to receive the laser-
induced acoustic signals. In addition, in many practical implementations of PAT the
photoacoustic signals are recorded over an aperture that does not enclose the object,
which results in a limited-angle tomographic reconstruction problem. In such cases,
the existing reconstruction algorithms, which are based on the least squares criteria
(i.e., the regularized Newton method) (see Chapters 1 and 2), often generate distorted
images with severe artifacts. Here we demonstrate that the total-variation-minimi-
zation (TVM)-based algorithm offers excellent photoacoustic image reconstruction
from few-detector and limited-angle data.
3.3.1 Mathematical Derivations
We describe the implementation of TVM within the FE-based reconstruction frame-
work in time-domain. To form an image from a presumably uniform initial guess of
the absorbed energy density distribution, we need a method of updating Ψ from its
starting value. This update is accomplished through the least squares minimization
of the following functional:
F ( p, Ψ) = ∑( p
j =1
o
j − pcj )2 , (3.9)

where poj and pcj are observed and computed acoustic field data for i = 1,2,…M
boundary locations. Using the regularized Newton method, we obtained the follow-
ing equation for updating Ψ:
(ℑT ℑ + λI) χ = ℑT ( po − pc ), (3.10)

where po = ( p1o , p2o ,… pMo )T , pc = ( p1c , p2c ,… pM ) , Δχ is the update vector for the
c T
absorbed optical energy density, ℑ is the Jacobian matrix formed by ∂ p/∂Ψ at the
boundary measurement sites, λ is the regularization parameter determined by com-
bined Marquardt and Tikhonov regularization schemes, and I is the identity matrix.
Two typical approaches exist for minimizing total variation: a constrained mini-
mization through the solution of the nonlinear photoacoustic (PA) equation (Rudin
et al. 1992; Paulsen and Jiang 1996) and an unconstrained minimization by addition
of the total variation as a penalty term to the least squares functional (Dobson and
Santosa 1994, 1996). From a computational standpoint, unconstrained minimiza-
tions are much easier to implement and require much fewer modifications to the
existing algorithm (Dobson and Santosa 1996). Thus here the unconstrained TVM
is used.
We now incorporate the total variation of Ψ as a penalty term by defining a new
functional (Yao and Jiang 2011):
F ( p, Ψ ) = F ( p, Ψ ) + L (Ψ ). (3.11)

Here

L (Ψ) =
∫ ω 2Ψ | Ψ |2 + δ 2 dx dy, (3.12)
is the penalty term, ω Ψ and δ are typically positive parameters that need to be deter-
mined numerically. The minimization of Equation (3.11) proceeds in standard fash-
ion by the differentiation of F with respect to each nodal parameter that constitutes
the Ψ distribution; simultaneously all these relations are set to zero. These steps lead
to the following nonlinear system of equations:
M
∂ F ∂ pcj
∂Ψ i
=− ∑( p
j =1
o
j − pcj )
∂Ψ i
+ Vi , (i = 1,2 N ), (3.13)

where
N N
∂ψ k ∂ψ i ∂ψ k ∂ψ i
∂L
ω 2Ψ ∑
k =1
Ψk
∂x ∂x
+ ∑Ψ
k =1
k
∂y ∂y
Vi =
∂Ψ i
=
∫ N
2
N
2
dx dy
∂ψ k ∂ψ k
ω 2Ψ ∑Ψk =1
k
∂x
+ ∑Ψ
k =1
k
∂y
+ δ2

Similarly to Equation (3.10), the following matrix equation for TVM constrained
inversion can be obtained
(ℑT ℑ + R + λI ) χ = ℑT ( po − pc ) − V , (3.14)

where
∂V1 ∂V1 ∂V1

∂Ψ1 ∂Ψ 2 ∂Ψ N
∂V2 ∂V2 ∂V2

R= ∂Ψ 1 ∂Ψ 2 ∂Ψ N

∂VN ∂VN ∂VN

∂Ψ1 ∂Ψ 2 ∂Ψ N

and V = {V1 ,V2 ,…VN }T .

3.3.2 Examples: Reduction of Noise Effect

In this section the TVM-enhanced reconstruction algorithm is tested and evalu-
ated using both simulated and experimental data for reduction of noise effect. For
comparative purposes, reconstructions without the TVM enhancement are also
presented.
3.3.2.1 Simulations
In these simulations, a dual meshing method, as detailed in Section 3.1, is used for
fast yet accurate inverse computation. The fine mesh used for the forward calcula-
tion consisted of 3627 nodes and 7072 elements, while the coarse mesh used for the
inverse calculation had 930 nodes and 1768 elements. All the images obtained from
the method without the TVM are the results of 3 iterations, while those obtained
from the TVM-enhanced method are the results of 15 or more iterations. As men-
tioned above, the parameters ωψ and δ were determined through numerical experi-
mentation. A constant value of δ = 0.001 was sufficient for the current simulation
and experimental studies, while the value of ωψ is related to the SNR of the mea-
surements (Yao and Jiang 2011). For the simulations presented, ωψ = 0.5 for the first,
second, and third cases, and ωψ = 1.0 for the fourth case were used.
For the first simulation, the test geometry is a 30 mm-diameter circular back-
ground containing an off-center 4 mm-diameter target region. The target had
Ψ = 2.0 mJ/mm 3, while the background had Ψ = 1.0mJ/mm 3 . In this case the image
reconstruction was performed with 0, 10, and 25% noise added to the “measured”
data. Figure 3.2 gives two sets of absorbed energy density images recovered using
the reconstruction method without (Figures 3.2a, 3.2c, and 3.2e) and with the TVM
enhancement (Figures 3.2b, 3.2d, and 3.2f) under the conditions of different noise
levels. As can be seen, enhancement of the reconstruction by the incorporation of
the TVM is obvious over the method without the TVM. To provide a more quantita-
tive assessment of these images, Figure 3.3 is included, in which the reconstructed
absorbed energy density profiles are displayed along transects through the target for
the images shown in Figure 3.2. We find that the TVM-enhanced method not only
can improve the quality of the recovered images but also can decrease the sensitivity
of the method to noise effect.
The second simulation is intended to investigate how the target size affects the
TVM enhancement. In this case, no noise was added to the “measured” data, and
the diameter of the off-center target was 2 mm. Figures 3.4a and 3.4b give two sets
of absorbed energy density images recovered using the method without the TVM
(Figure 3.4a) and with the enhancement (Figure 3.4b), while Figures 3.4c and 3.4d
present the quantitative profiles of the absorbed energy density along transects
through the target for the images shown in Figures 3.2a,b and Figures 3.4a,b. Again,
considerable improvement can be observed from the reconstructed results when the
TVM is invoked compared with the method without the TVM. It is also interesting
to note that the enhancement for this case is more striking than that for the first one
where the background contained a larger target.
The third simulation aims to see how the contrast level of the absorbed energy
density between the target and the background impacts the TVM enhancement.
10 2 10 2
1.8
5 5
1.6
0 1.5 0
1.4
–5 –5
1.2
–10 1 –10
1
–10 0 10 –10 0 10
(a) (b)
2 10 2
10
1.8
5 5
1.6
0 1.5 0
1.4
–5 –5
1.2
–10 1 –10
1
–10 0 10 –10 0 10
(c) (d)
2.2
10 2 10 2
5 5 1.8
0 1.6
1.5 0
1.4
–5 –5
1.2
–10 1 –10
1
–10 0 10 –10 0 10
(e) (f )
FIGURE 3.2 Reconstructed absorbed energy density images from simulated data with and
without the total-variation-minimization (TVM) enhancement under different noise levels
(case 1). (a) Without the TVM, 0% noise. (b) With the TVM, 0% noise. (c) Without the TVM,
10% noise. (d) With the TVM, 10% noise. (e) Without the TVM, 25% noise. (f) With the
TVM, 25% noise. The axes (left and bottom) illustrate the spatial scale in mm, whereas the
color scale (right) records the absorbed energy density in mJ/mm3.
In this case, no noise was added to the “measured” data, and the off-center tar-
get had a diameter of 4 mm and Ψ = 1.5 mJ/mm 3 . Figures 3.5a and 3.5b present
two sets of absorbed energy density images recovered using the method with-
out the TVM (Figure 3.5a) and with the TVM enhancement (Figure 3.5b), while
Figures 3.5c and 3.5d show the quantitative profiles of the absorbed energy density
Exact Exact
Without TVM Without TVM
Absorbed Energy Density

2.0 2.0
With TVM With TVM
1.5 1.5
1.0 1.0
–10 –5 0 5 10 –10 –5 0 5 10
X Position/mm X Position/mm
(a) (b)
Exact
Without TVM
2.0
With TVM
1.5
1.0
–15 –10 –5 0 5 10 15
X Position/mm
(c)
FIGURE 3.3 Comparison of the exact and reconstructed absorbed energy density profiles
along transect y = 0 mm for the images appearing in Figure 3.2. (a) 0% noise. (b) 10% noise.
(c) 25% noise.
along transects through the target for the images shown in Figures 3.2a,b and
Figures 3.5a,b. We can see that the images formed by incorporation of the TVM
are clearly enhanced qualitatively in visual content relative to that obtained using
the method without the TVM. We also note that the lower the contrast level, the
larger the enhancement.
In the fourth simulation, three targets having different shapes (circular, ellipti-
cal, and rectangular) were embedded in the background, and the absorbed energy
density of these targets was 2.5, 1.5, and 2.0 mJ/mm 3 , respectively. A noise level of
25% was added to the “measured” data in this case. Figure 3.6 shows the exact and
the recovered absorbed energy density images as well as the quantitative absorbed
energy density property profiles along the transect that across these targets. Again,
the improvement in image quality is apparent.
To quantitatively evaluate the reconstruction quality using the reconstruction
method with and without TVM enhancement, we use the universal quality index
2.2
2.2
10 10 2
2
5 1.8 5 1.8
0 1.6 0 1.6
–5 1.4 –5 1.4
1.2 1.2
–10 –10
1 1
–10 0 10 –10 0 10
(a) (b)
Exact 2.5 Exact


2.0 With TVM With TVM
2.0
1.5
1.5
1.0 1.0
2.5 5.0 7.5 10.0 12.5 2.5 5.0 7.5 10.0 12.5
X Position X Position/mm
(c) (d)
without the total-variation-minimization (TVM) enhancement with different target sizes
(case 2). (a) Without the TVM, 2 mm-diameter target. (b) With the TVM, 2 mm-diameter
target. (c) The absorbed energy density profiles along the transect y = 0 mm for the images
appearing in Figures. 3.2a and 3.2b (4 mm-diameter target). (d) The absorbed energy density
property profiles along the transect y = 0 mm for the images appearing in Figures 3.4(a) and
3.4(b) (2 mm-diameter target). For (a) and (b), units of measurements on the axes are in mm
and units of measurements for numbers given at the color bar are in mJ/mm3.
(UQI) to measure the degree of similarity between the reconstructed and exact
images (Wang and Bovik 2002). We first interpreted an image f as vectors of size
N: f = ( f1 , f2 ,… f N )T , where superscript T denotes the matrix transpose, and N, the
number of image data acquired from the FE-based algorithm. We then defined image
means, variances, and covariances over the whole image domain as
N
1
fj=
N ∑ f , (3.15) k
j
k =1
N
1

j
σ =
N ∑( f
k =1
k
j
−f j 2
) , (3.16)
1.6
1.5
10 10
1.4 1.4
5 5
1.3
0 0
1.2 1.2
–5 –5
1.1
–10 1 –10
1
–10 0 10 –10 0 10
(a) (b)
Exact Exact

2.0 With TVM With TVM

1.5
1.5
1.0
1.0
2.5 5.0 7.5 10.0 12.5 2.5 5.0 7.5 10.0 12.5
X Position X Position/mm
(c) (d)
without the total-variation-minimization (TVM) enhancement with different contrast levels
between the target and the background (case 3). (a) Without the TVM, 1.5:1 contrast. (b) With
the TVM, 1.5:1 contrast. (c) Absorbed energy density profiles along the transect y = 0 mm for
the images appearing in Figures. 3.2(a) and 3.2(b) (2:1 contrast). (d) Absorbed energy density
profiles along the transect y = 0mm for the images appearing in Figures 3.5(a) and 3.5(b)
(1.5:1 contrast). For (a) and (b), units of measurements on the axes are in mm and units of
measurements for numbers given at the color bar are in mJ/mm3).
where j = 0 and 1, and

N
1
Cov{f 1 , f 0 } =
N ∑( f 1
k − f1 )( f k
0
)
− f 0 , (3.17)
k =1
Hence, the UQI is defined as
2Cov{f 1 , f 0 } 2f1f 0
UQI{f 1 , f 0 } = . (3.18)
(σ1 ) 2 + (σ 0 ) 2 ( f 1 ) 2 + ( f 0 ) 2
UQI measures the image similarity between the reconstructed (f 1) and reference/
exact (f 0) images, and its value ranges between 0 and 1. The value of the UQI is
2.4
10 2.2 10 2.5
5 2 5
2
0 1.8
0
1.6 1.5
–5 –5
1.4
–10 1.2 –10 1
–10 0 10 –10 0 10
(a) (b)
3.0
Exact

2.5 Without TVM
10 2.5
With TVM
5 2 2.0
0
1.5
–5 1.5
1.0
–10
1
0.5
–10 0 10 –15 –10 –5 0 5 10 15
X Position/mm
(c) (d)
without the total-variation-minimization (TVM) enhancement for three targets having dif-
ferent shapes (case 4). (a) Exact image. (b) Without the TVM. (c) With the TVM. (d) The
absorbed energy density profiles along the transect y = 0 mm. For (a), (b), and (c), units of
measurements on the axes are in mm and units of measurements for numbers given at the
color bar are in mJ/mm3).
closer to 1 when the reconstructed image is more similar to the exact image. We cal-
culated the UQIs for the simulation cases presented above, and the results are given
in Figure 3.7 where Figures 3.7a and 3.7b show the UQIs for the first case when a
different noise level is considered and for all four cases without added noise. In
Figure 3.7, the horizon axis indicates the case number (from cases 1 to 4, presented
above), and the vertical axis shows the value of UQI. We immediately note from
Figure 3.7 that the TVM-based method provides significantly better UQIs than that
without TVM.
3.3.2.2 Phantom Experiments
In this section phantom experimental data were used to confirm our findings from
the simulations. The experimental setup used for collecting the phantom data
was a pulsed ND: YAG laser-based single transducer (1 MHz) scanning system
described in Chapter 1. Three phantom experiments were conducted. In the first two
1.00 1.00
0.98
0.98
0.96
0.94 0.96
UQI
UQI
0.92
0.94
0.90
0.88 Without TVM 0.92 Without TVM
With TVM With TVM
0.86
0.90
0% 10% 25% 1 2 3 4
Noise Level Case #
FIGURE 3.7 UQI calculated from the recovered images with and without TVM enhance-
ment from simulated data. (a) Case 1 when different noise levels are considered. (b) Cases 2–4.
experiments, we embedded one or two objects with a size ranging from 3.0 to 0.5mm
in a 50mm-diameter solid cylindrical phantom. The phantom materials used con-
sisted of Intralipid as scatterer and India ink as absorber with Agar powder (1–2%)
for solidifying the Intralipid and India ink solution. The absorption coefficient of the
background phantom was 0.01 mm–1, while the absorption coefficient of the target(s)
was 0.03 mm–1. In the last experiment, we used a single-target-containing phantom,
aiming to test the capability of detecting target having low optical contrasts relative
to the background phantom. In this case, the target had an absorption coefficient of
0.015 mm–1. The reduced scattering coefficients of the background phantom and
targets were 1.0 and 3.0 mm–1 for the first two experiments, and 1.0 and 2.0 mm–1 for
the last experiment.
A total of 120 receivers were equally distributed along the surface of the circular
background region. In the reconstructions, the fine mesh used for the forward cal-
culation consisted of 5977 nodes and 11,712 elements, while the coarse mesh used
for the inverse calculation had 1525 nodes and 2928 elements. The reconstructed
images were the results of 3 and 15 iterations for the method without and with the
TVM, respectively. For these experimental cases, ω Ψ = 1.0 and δ = 0.001 for the
first and third cases and ω Ψ = 2.0 and δ = 0.001 for the second case were used.
Here we note that in the experimental situation, the effect of ultrasonic transducer
response should be considered in the PAT image reconstruction because the trans-
ducer mechanical-electrical impulse response as well as the transducer aperture
effect may introduce significant errors in the forward model. To reduce these effects
in our calculations, we applied the normalized acoustic pressure distribution in the
reconstruction and used the appropriate initial parameters obtained from a search-
ing method (Jiang 2010).
Figure 3.8 shows the reconstructed absorbed energy density images for all the
three experimental cases, and Figure 3.9 presents quantitative absorption coefficient
profiles along transects through one target for the images shown in Figure 3.8. We
see that considerably enhanced images are achieved with the TVM, especially when
1.5
20 20
1 1
10 10
0 0.5 0
0.5
–10 –10
0
0
–20 –20
–20 –10 0 10 20 –20 –10 0 10 20
(a) (b)
1 1
20 20
0.8
10 0.5 10
0.6
0 0 0.4
0
–10 –10 0.2
0
–20 –0.5 –20
–20 –10 0 10 20 –20 –10 0 10 20

(c) (d)
1
20 20
0.8 0.8
10 0.6 10 0.6
0 0.4 0 0.4
0.2
–10 0.2
–10
0
0
–20 –0.2 –20
–20 –10 0 10 20 –20 –10 0 10 20

(e) (f )
FIGURE 3.8 Reconstructed absorbed energy density images from the three phantom exper-
iments. (a) Case 1 without the TVM. (b) Case 1 with the TVM. (c) Case 2 without the TVM.
(d) Case 2 with the TVM. (e) Case 3 without the TVM. (f) Case 3 with the TVM. Units of
measurements on the axes are in mm and units of measurements for numbers given at the
color bar are mJ/mm3.
Without TVM 1.2 Without TVM

0.8

With TVM With TVM
0.8
0.4
0.4
0.0
0.0
–0.4
–0.4
–20 –10 0 10 20 –20 –10 0 10 20
(a) (b)

With TVM
0.4 With TVM
0.8
0.2
0.4
0.0
0.0
–0.2
–20 –10 0 10 20 –20 –10 0 10 20
(c) (d)
FIGURE 3.9 Recovered absorbed energy density profiles along (a) y = –7.0 mm crossing the
3 mm-diameter target for experimental case 1, (b) y = 8.0 mm crossing the 2 mm-diameter
target for experimental case 1, (c) y = 1.0 mm for experimental case 2, and (d) y = 6.5 mm for
experimental case 3.
the target is small (case 2), or when the contrast level between the target and the
background is low (case 3).
Because there is no true or exact image available for the experimental cases, we
use the contrast-to-noise ratio (CNR) to quantitatively evaluate the experimental
results. To compute CNR, we selected two regions of interest (ROIs) (each has the
same size). One region is in the target (referred as t-ROI) and the other is in the back-
ground (denoted as b-ROI). The CNR is defined as
f (t ) − f ( b )
CNR = , (3.19)
σ(b)
where f (t ) and f ( b ), respectively, denote the mean over the t-ROI and b-ROI, and σ( b )
denotes the variances over the b-ROI. We used solid- and dashed-line circles to express
1
20 20
1 0.8
10 1 10
0.6
0 0.5 0 3 0.4
2
–10 –10 0.2
0 0
–20 –20
–20 –10 0 10 20 –20 –10 0 10 20
(a) (b)
30
Without TVM
20 0.8 With TVM
10 0.6 20
4
CNR
0 0.4
0.2 10
–10
0
–20
0
–20 –10 0 10 20 1 2 3 4
Region #
(c) (d)
FIGURE 3.10 CNR calculated for the recovered images using the method with and without
TVM. (a)–(c) Images shown in Figures. 3.8(b), 3.8(d), and 3.8(f) with the ROIs marked (four
pairs of ROIs; solid-line-circle: t-ROI; dashed-line-circle: b-ROI). (d) CNR computed for the
4 pairs of ROIs shown in (a)–(c). For (a), (b), and (c), units of measurements on the axes are
mm and units of measurements for numbers given at the color bar are mJ/mm3.
the t-ROI and b-ROI for the three experimental cases. We calculated the CNR for each
pair of the t- and b-ROIs using Equation (3.19) and present the results in Figure 3.10.
We see that the CNR obtained using the method with TVM is clearly larger than that
using the method without TVM, indicating that the TVM-based method is less sensi-
tive to the noise effects.
3.3.3 Examples: Image Reconstruction from Few-

Detector and Limited-Angle Data
In this section our TVM enhanced reconstruction algorithm is tested and evaluated
using phantom experimental data for image reconstruction from few-detector and
limited-angle data. In the first two experiments, we embedded one or two objects
with a size ranging from 3 to 0.5 mm in a 50 mm-diameter solid cylindrical phan-
tom. The absorption coefficient of the background phantom was 0.01 mm–1, while
the absorption coefficient of the target(s) was 0.03 mm–1. In the last experiment, we
used a single-target-containing phantom, aiming to test the capability of detecting

target having low optical contrasts relative to the background phantom. In this case,
the target had an absorption coefficient of 0.015 mm–1. The reduced scattering coeffi-
cients of the background phantom and targets were 1.0 and 3.0 mm–1 for the first two
experiments, and 1.0 and 2.0 mm–1 for the last experiment. In the reconstructions,
we used the dual-meshing scheme where the fine mesh used for the forward calcula-
tion consisted of 5977 nodes and 11,712 elements, while the coarse mesh used for the
inverse calculation had 1525 nodes and 2928 elements. All the images obtained from
the method without the TVM are the results of 3 iterations, while those obtained
from the TVM-enhanced method are the results of 20 or more iterations. In the
reconstruction, ω Ψ = 1.0 and δ = 0.001 for the first case, and ω Ψ = 0.5 and δ =
0.001 for the second and third cases were used.
The reconstruction results based on few-detector data from the three experimen-
tal cases are shown in Figures 3.11, 3.12, and 3.13, respectively. In each figure, the
images in the top and bottom rows, respectively, present the recovered absorbed
energy density images using the algorithm without and with the TVM where 120,
60, 30, and 15 detectors were equally distributed along the surface of the circu-
lar background region. As expected, the conventional algorithm without the TVM
provides high-quality images only when the number of detectors was relatively
sufficient (Figures 3.11a,b, Figures 3.12a,b, and Figures 3.13a,b). The images recon-
structed from few-detector data by the conventional algorithm without the TVM
contained severe artifacts and distortions (Figures 3.11c,d, Figures 3.12c,d, and
Figures 3.13c,d). Considerably enhanced images are achieved using the method
with the TVM, especially when the number of the detectors is reduced to 30 or 15
(Figures 3.11g,h, Figures 3.12g,h, and Figures 3.13g,h). We also note that the com-
putational efficiency of our TVM-based algorithm for recovering a small absorber
(Figure 3.12) and a larger absorber (Figure 3.13) is similar when the number of the
detectors is insufficient.
Images reconstructed from the limited-angle data for the first case using the two
methods are displayed in Figure 3.14, where the top and bottom rows, respectively,
show the recovered absorbed energy density images using the algorithm without
and with the TVM when 120 detectors over 360°, 60 detectors over 180°, and 30
detectors over 90° were equally distributed along the surface of the circular back-
ground region. Again, strong artifacts and distortions exist in the images recov-
ered with the method without the TVM when the data collected are angle-limited
(Figures 3.14b,c), while considerably improved images are reconstructed using the
TVM-enhanced algorithm (Figures 3.14e,f).
We calculated the UQIs for the three cases presented above, and the resuls are
given in Figure 3.15. Figures 3.15a and 3.15b show the results from the few-detector
data based on the method without and with the TVM, respectively, and Figure 3.15c
presents the results from the limited-angle data for the first case based on the two
methods. Here the image data recovered by 120 detectors was regarded as the refer-
ence one. The computed UQIs shown in Figure 3.15 confirm the observation that
the TVM-enhanced PAT algorithm provides significantly better image quality com-
pared to the conventional method without the TVM.
20 20
1 1
10 10
0 0.5 0 0.5
–10 –10
0 0
–20 –20
–20 –10 0 10 20 –20 –10 0 10 20

(a) (b)
1
20 20 0.8
0.8
10 10 0.6
0.6
0.4
0 0.4 0
0.2 0.2
–10 –10 0
0
–20 –0.2 –20 –0.2
–20 –10 0 10 20 –20 –10 0 10 20

(c) (d)
20 20
1 1
10 10
0 0.5 0
0.5
–10 –10
0 0
–20 –20
–20 –10 0 10 20 –20 –10 0 10 20
(e) (f )
1
20 1 20
0.8
10 0.8 10
0.6 0.6
0 0
0.4 0.4
–10 0.2 –10 0.2
–20 0 –20 0
–20 –10 0 10 20 –20 –10 0 10 20

(g) (h)
FIGURE 3.11 Reconstructed photoacoustic images based on few-detector data for case 1.
(a) 120 detectors, without TVM. (b) 60 detectors, without TVM. (c) 30 detectors, without
TVM. (d) 15 detectors, without TVM. (e) 120 detectors, with TVM. (f) 60 detectors, with
TVM. (g) 30 detectors, with TVM. (h) 15 detectors, with TVM.
20 20 0.6
10 0.5 10 0.4
0 0 0.2
0 0
–10 –10
–0.2
–20 –0.5 –20
–0.4
–20 –10 0 10 20 –20 –10 0 10 20
(a) (b)
0.6 0.5
20 20
0.4
10 10
0.2
0 0 0 0
–10 –0.2 –10
–20 –0.4 –20

–0.5
–20 –10 0 10 20 –20 –10 0 10 20
(c) (d)
1
20 20 0.8
0.6
10 0.5 10
0.4
0 0 0.2
0 0
–10 –10
–0.2
–20 –0.5 –20 –0.4
–20 –10 0 10 20 –20 –10 0 10 20

(e) (f )
20 20 0.6
0.6
10 10 0.4
0.4
0 0 0.2
0.2
–10 –10 0
0
–0.2
–20 –0.2 –20
–20 –10 0 10 20 –20 –10 0 10 20
(g) (h)
1
20 20
0.8 0.8
10 0.6 10 0.6
0 0.4 0 0.4
0.2 0.2
–10 –10
0 0
–20 0.2 –20
–0.2
–20 –10 0 10 20 –20 –10 0 10 20
(a) (b)
1
20 20 0.8
0.6
10 10
0.5 0.4
0 0 0.2
0
–10 0 –10
–0.2
–20 –20 –0.4
–20 –10 0 10 20 –20 –10 0 10 20
(c) (d)
1
20 20
0.8 0.8
10 10 0.6
0.6
0 0.4 0 0.4
0.2 –10 0.2
–10
0 0
–20 –20
–0.2
–20 –10 0 10 20 –20 –10 0 10 20
(e) (f )
20 20 0.8
0.8
10 0.6 10 0.6
0 0.4 0 0.4
–10 0.2 –10 0.2
–20 0 –20 0
–20 –10 0 10 20 –20 –10 0 10 20

(g) (h)
1.5
20 20
1 1
10 10
0 0.5 0 0.5
–10 –10
0 0
–20 –20
0.5
–20 –10 0 10 20 –20 –10 0 10 20
(a) (b)
20 20
1
10 0.5 10
0 0 0.5
0
–10 –10
0
–20 –0.5 –20
–20 –10 0 10 20 –20 –10 0 10 20

(c) (d)
1.5
0.8
20 20
1 0.6
10 10
0.4
0 0.5 0
0.2
–10 0 –10 0
–20 –20 –0.2

–0.5
–20 –10 0 10 20 –20 –10 0 10 20
(e) (f )
FIGURE 3.14 Reconstructed photoacoustic images based on limited-angle data for case 1.
(a) 120 detectors over 360°, without TVM. (b) 60 detectors over 180°, without TVM. (c) 30
detectors over 90°, without TVM. (d) 120 detectors over 360°, with TVM. (e) 60 detectors
over 180°, with TVM. (f) 30 detectors over 90°, with TVM.
3.4 RADIATIVE TRANSFER EQUATION

All the quantitative PAT (qPAT) methods described in Chapter 2 are based on the
photon diffusion equation (DE), the first-order approximation to the rigorous radiative
transfer equation (RTE). While the DE is widely used for describing light propagation
1.00 1.00
0.75 0.75
UQI
UQI
0.50 0.50
Experiment case 1 Experiment case 1

0.25 Experiment case 2 0.25 Experiment case 2
Experiment case 3 Experiment case 3
0.00 0.00
120 90 60 30 120 90 60 30
Detector Numbers Detector Numbers
(a) (b)
1.00
0.75
UQI
0.50
0.25
W/o TVM
W/ TVM
0.00
360° 180° 90°
Scanning Degree
(c)
FIGURE 3.15 UQIs calculated from the recovered images for the three cases based on few-
detector data without TVM (a) and with TVM (b) and for case 1 based on limited angle data
with and without TVM.
in tissue, it has well-known limitations (Firbank et al. 1996; Hielscher et al. 1998;
Dehghani et al. 1999). For example, it fails to describe accurately light propagation
in medium where its absorption coefficient is not much smaller than the scattering
coefficient, or in so-called void-like regions where the absorption and scattering coef-
ficients are very low. Another situation where the diffusion approximation is not accu-
rate enough is in modeling light propagation in media of small volume, because of
the small optical distance between sources and measurements (Ren et al. 2007). In
addition, the variable anisotropy factor, g, also affects the photon migration in tissue,
which is not fully considered in the DE (Klose et al. 2002). In the area of diffuse
optical tomography (DOT), it has been shown that improved reconstruction of optical
properties can be obtained using RTE or higher-order diffusion equations (Klose et
al. 2002; Xu et al. 2003; Ren et al. 2007; Yuan et al. 2009).
Here we describe a RTE-based qPAT method that allows for quantitative recov-
ery of absorption coefficient images of heterogeneous media using tomographic
photoacoustic measurements. In this method, acoustic measurements from the con-
ventional PAT are combined with the RTE to separate the product of absorption
coefficient and optical fluence. We demonstrate our reconstruction algorithm using
considerable simulated data under various practical scenarios and validate it using a
series of tissue-like phantom experiments.
3.4.1 The RTE and Its Finite Element Discretization

Transport theory is usually applied to describe transport of energy through a medium
containing scattering particles, such as light propagation in biological tissue. The
RTE is a one-speed approximation of the transport equation, and it is obtained by
integrating the transport equation over energy and by approximating the parame-
ters by energy-averaged parameters (Case and Zweifel 1967; Duderstadt and Martin
1979). The RTE in the steady-state can be described as

( ⋅ + s + a )φ(r , ) = s
∫ S n −1
φ(r , )Θ( , ) d + q(r , ), (3.20)

where s = s (r ) and a = a (r ) are the scattering and absorption coefficients of

the
medium, respectively; φ (r , ) is the radiance; and q (r , ) is the source term;
∈ S n−1 denotes a unit vector in the direction of interest. Here S n−1 is the angular
direction; n = 2 or 3 denotes the physical domain, which is considered isotropic in
the sense that the probability of scattering between two directions depends only on
the relative angle between those directions. R is the spatial domain, and ∂R denotes
its boundary. The kernel Θ( , ) is the scattering phase function describing the
probability
density that a photon with an initial direction will have a direc-
tion after a scattering event. Here we assume that the scattering phase function
depends only on the angle between the incoming and outgoing directions, and thus

Θ( , ) = Θ( ⋅ ). (3.21)
The 2D Henyey–Greenstein scattering function, the most widely adopted and

highly accurate phase function for isotropic medium, is used here (Henyey and
Greenstein 1941; Heino et al. 2003):
1 1 − g2
Θ( ⋅ ) = 2
. (3.22)
2π (1 + g − 2 g cos γ )

where γ is the angle between the input direction and output direction . The
anisotropy factor, g (−1 < g < 1), defines the shape of the probability density. It is
known that g is between 0.7 and 0.99 for most tissues (Wnek and Bowlin 2004).
The boundary conditions (BC) for the RTE assume that no photons travel in an
inward direction at the boundary ∂R, that is,

φ(r , ) = 0, ⋅ nˆ < 0 for all r ∈∂ R (3.23)

where n̂ is the outward unit normal on ∂S. This BC, also known as free surface BC
or vacuum BC, implies that once a photon escapes the domain, it does not reenter it.

The BC can be modified to include a boundary source φ0 (r , ) at the source posi-
tion εi ∂R, and can be written as follows (Tervo et al. 1999):

φ 0 (r , ), r ∈∪ i εi ⋅ nˆ < 0
φ(r , ) = (3.24)
0, r ∈∂ S \∪i ε i ⋅ nˆ < 0

The solution to the RTE with the above boundary conditions is existent and unique
(Duderstadt and Martin 1979; Qatanani et al. 2007). To obtain the Galerkin weak form
of the RTE, Equation (3.20) is multiplied by a set of locally spatially varying Lagrangian

test/basis function, ψ j = ψ j (r , ) and integrated over the physical domain, R and the
angular directions, S n−1, yielding the finite-element discretization of Equation (3.20):
N

− ∑φ ∫ ∫
j =1
j
R S n −1
ψ j (r , ) ⋅ ψ i (r , ) d dr
N

+ ∑ φ ∫ ∫
j =1
j
∂R S n −1
( ⋅ nˆ )+ ψ j (r , )ψ i (r , ) d dS
N

+ ∑φ ∫ ∫
j =1
j
R S n −1
( a + s )ψ j (r , )ψ i (r , ) d dr (3.25)
N

− ∑φ ∫ ∫
j =1
j
R S n −1
s
∫ S n −1
Θ( ⋅ )φ(r , ) d ψ i (r , ) d dr
N

= ∑ φ ∫ ∫
j =1
0
j
∂R S n −1
( ⋅ nˆ )− ψ j (r , )ψ i (r , ) d dS

+
∫∫ R s n −1
ψ i (r , )q(r , ) d dr

where φ0j (r , ) is the source
term at the boundary, ( ⋅ nˆ ) ± denotes the positive

and the negative parts of ( ⋅ nˆ ) , and q(r , ) is the actual light source numerically
placed inside the domain.

In addition, φ(r , ) needs to be expanded as the sum of coefficients multiplied by
the Lagrangian basis function:
N Nn Na

φ(r , ) = ∑ j =1
ψ j (r , )φ j = ∑∑ψ
nj =1 mj =1
nj (r )ψ mj ( )φnj ,mj (3.26)

where ψ nj (r ) and ψ mj ( ) are the nodal spatial and angular basis functions; φnj ,mj is
the radiance at the spatial nodal point, nj, and direction, mj; Nn is the number of spa-
tial nodes of the mesh; and Na is the number of angular directions. It is known that the
ray effect may disturb the standard FE-techniques when solving the RTE, since it can
produce oscillating results, or it can visually be seen as “photon rays” radiating from
the source into the direction of the discretization angles (Kanschat 1998; Tarvainen
et al. 2005). To overcome the ray effect, the streamline diffusion modification (SDM)
is used in theFE-solution of the RTE. In the SDM, the test function is of the form

(ψ j (r , ) + δ ⋅ ψ j (r , )) instead of the standard form of a test function (ψ j (r , )).
The parameter δ is the “smoothing” parameter, which is a spatially varying constant
that depends on the local absorption and scattering (Kanschat 1998).
Equation (3.25) can be rewritten in matrix form:
[ K1 + K1,sdm + K 2 + K 3 + K 3,sdm + K 4 + K 4,sdm ]{φ} = [ B1 ]{φ0 } + {B2} (3.27)

where
2π
∂ψ ni (r ) ∂ψ (r )

K1ij = −
∫ R
ψ nj (r )
∫ 0
cos θ
∂x
+ sin θ ni
∂y
ψ mj (θ)ψ mi (θ) d θ dr

2π
∂ψ ni ( r ) ∂ψ ( r ) ∂ψ nj ( r ) ∂ψ ( r )
K1,ijsdm =
∫ ∫ R
δ
0
cos θ
∂x
+ sin θ ni
∂y
cos θ
∂x
+ sin θ nj
∂y

ψ mj ( θ ) ψ mi ( θ ) dθdr

2π

K 2ij =
∫ ∂R
ψ nj (r )ψ ni (r )dS
∫ 0
(nˆ x cos θ + nˆ y sin θ)+ ψ mj (θ)ψ mi (θ) d θ
2π

K 3ij =
∫ R
( a + s )ψ nj (r )ψ ni (r ) dr
∫ 0
ψ mj (θ)ψ mi (θ) d θ
2π
∂ψ ni (r ) ∂ψ (r )
K 3,ij sdm =
∫ R
δ( a + s )ψ nj (r ) dr
∫0
cos θ
∂x
+ sin θ ni
∂y
ψ mj (θ)ψ mi (θ) d θ
2π 2π

K 4ij = −
∫ R
s ψ nj (r )ψ ni (r )dr
∫0
ψ mi (θ) d θ
∫0
ψ mj (θ )Θ(θ ⋅ θ ) d θ
2π
∂ψ ni (r ) ∂ψ (r )
K 4,ij sdm = −
∫ R
δ s ψ nj ( r )dr
∫ 0
cos θ
∂x
+ sin θ ni
∂y
ψ mi ( θ ) d θ
2π

∫0
ψ mj (θ )Θ(θ ⋅θ ) d θ
2π

B1ij =
∫ ∂R
ψ nj ( r ) ψ ni ( r )dS
∫ 0
(nˆ x cos θ + nˆ y sin θ)− ψ mj (θ)ψ mi (θ) d θ
2π

B2i =
∫∫
R 0
q(r , θ)ψ mi (θ) d θψ ni (r )dr

{φ} = {φ1 , φ2 ,φ N n N a }T ;{φ0 } = φ10 , φ20 ,φ0N n N a { }
where the outward unit normal is denoted by nˆ = (nˆ x , nˆ y ), S n−1 = S 1 the unit cir-
cle, = (cos θ,sin θ) a point on S 1 , i = N a (ni − 1) + mi , j = N a (nj − 1) + mj , and
(i, j = 1,… N n N a ) .
3.4.2 The RTE-Based Quantitative PAT

The reconstruction method for the RTE-based qPAT includes two steps. The first is
to obtain the map of absorbed optical energy density through the FE-based recon-
struction algorithm described in Chapter 1. The second step is to recover the distri-
bution of absorption coefficient from the absorbed energy density obtained in the
first step based on the finite element solution to the RTE. Since our reconstruction
algorithm for the first step has been described in detail in Chapter 1, we give a
brief outline here in order to couple the second step with the first step. The core
procedure of our PAT reconstruction algorithm can be described by the following
two equations:
2 v0βΨ(r )
p(r , ω ) + k02 p(r , ω ) = ik0 ; (3.28)
Cp

(ℑT ℑ + λI) χ = ℑT ( po − pc ) (3.29)

where p is the pressure wave; k0 = ω/v0 is the wave number described by the angu-
lar frequency, ω, and the speed of acoustic wave in the medium, v0; β is the ther-
mal expansion coefficient; Cp is the specific heat; Ψ is the absorbed energy density
that is product of absorption coefficient, μa, and optical fluence, Φ (i.e., Ψ = a Φ );
po = ( p1o , p2o ,… pMo )T , pc = ( p1c , p2c ,… pM ) , and pio and pic are observed and com-
c T
puted complex acoustic field data for i = 1,2,⋯M boundary locations; Δχ is the update
vector for the absorbed optical energy density; ℑ is the Jacobian matrix formed by
∂ p/∂Ψ at the boundary measurement sites; λ is the regularization parameter deter-
mined by combined Marquardt and Tikhonov regularization schemes; and I is the
identity matrix. Thus here the image formation task is to update absorbed optical
energy density distribution via iterative solution of Equations (3.28) and (3.29) so that
an object function composed of a weighted sum of the squared difference between
computed and measured acoustic data can be minimized.
The second step is to determine the μa distribution by the following iterative solu-
tion procedure using finite element method (see Section 2.1): (a) choose an initial
value for μa, e.g., a = 0.0001 mm −1; (b) compute the optical fluence Φ using the
RTE; (c) calculate the absorbed energy density by Ψ c = a Φ ; (d) compute the error
between the measured Ψ (from the first step) and calculated Ψ c , and μa is updated
using a = Ψ/Φ; and (e) if the error is sufficiently small, then the iterative calcula-
tion stops; otherwise repeat steps (b) to (d) until a small error is reached. This pro-
cedure converges within about 10 to 20 iterations generally. In the reconstruction,
Source
FIGURE 3.16 Test geometry used for the simulations described in Section 3.4.3.1.
μs, and g are assumed as constant, while the incident laser source strength and the
absorbed energy density Ψ are estimated in advance.
3.4.3 Results and Discussion

3.4.3.1 Comparison of the Transport and Diffusion Calculations
In this subsection we perform comparative simulations using the RTE and DE. The
test geometry is schematically shown in Figure 3.16 where a centered circular target
of 20 mm diameter is embedded in a 100 × 100 mm box/background. An isotropic
source is positioned at the center of the left side of the box and vacuum boundary
conditions are applied to the other three sides of the box. Several simulations were
conducted where the central target had weak, moderate, and strong absorption and
very low absorption and scattering. A mesh consisting of 4067 nodes and 7932 tri-
angular elements was used in the calculations. In this study we typically used 32
angular directions for the calculations (we found that larger numbers of directions
changed the solutions insignificantly by less than 0.5%). The “smoothing” parameter
in the streamline diffusion modification was chosen as δ = a 1+ s .
For the first and second simulations, the target had a = 0.1 mm −1 and a = 1.0 mm −1,
respectively, and the background had a = 0.01 mm −1, while s = 1.0 mm −1(reduced
scattering coefficient) for the target and background for both cases. For the third
simulation, the target had s = 0.01 mm −1 and a = 0.0001 mm −1which are much
smaller than that for the background ( s = 0.5 mm −1 and a = 0.005 mm −1 ). g = 0
for all three simulations. The calculated photon fluxes are plotted along a horizontal
line passing through the center of the target in Figure 3.17a for the first and second
simulations and in Figure 3.17b for the third simulation. We see from Figure 3.17a
that the RTE and DE calculations agree well throughout the media for the first
simulation. However, when a is increased inside the target to be comparable to
s for the second simulation, large differences between the two calculations are
100
µa = 0.1 mm–1
Photon Intensity (Photons/mm2)

1
0.01
µa = 1.0 mm–1
1E–4
1E–6
1E–8 RTE
DE
1E–10
20 30 40 50 60 70 80
Position (mm)
(a)
Photon Intensity (Photons/mm2)
0.1
0.01 RTE
DE
20 30 40 50 60 70 80
Position (mm)
(b)
FIGURE 3.17 Comparison of the diffusion and transport calculations: (a) influence of
μa /μ′ ;
s (b) influence of an almost scattering- and absorption-free heterogeneity.
observed inside the target, while the two models agree well outside the target
region. We also observe significant differences between the two calculations for
the third simulation although the diffusion approximation holds everywhere in the
media for this case. We note that all these results shown in Figure 3.17 agree well
with those published in the literature involving transport calculations (Klose et al.
2002, Aydin et al. 2004).
FIGURE 3.18 Test geometry used for the simulations described in Section 3.4.3.2.1.
3.4.3.2 Comparison of the DE- and RTE-Based qPAT

3.4.3.2.1 Simulations
We present considerable simulations to evaluate the performance of both the DE-
and RTE-based qPAT reconstructions. Our goal here is to quantify the errors of
the DE-based method in recovering absorption coefficient in situations where the
DE is not an accurate approximation to the RTE. In these simulations, the forward
“measured” data were generated using the RTE coupled with the photoacoustic
wave equation. A mesh consisting of 5081 nodes and 9920 triangular elements
along with 64 angular directions was used for this purpose. The test geometry is
schematically shown in Figure 3.18 where a 30 mm-diameter circular background
contains a 6 mm-diameter central target. For all the simulations presented, the
background had: a = 0.01 mm −1 and the target had: a = 0.02 mm −1. We consider
three groups of simulations aiming to study the effects of a / s ratio, g factor and
imaging domain size on qPAT reconstruction. A mesh consisting of 1301 nodes and
2480 triangular elements along with 64 angular directions was used in the calcula-
tions. For the first step of recovering the absorbed energy density, each iteration
costed about 10 minutes, while for the second step of obtaining the absorption
coefficient, each iteration costed about several seconds for the DE-based algorithm
and about 30 minutes for the RTE-based algorithm on a 2.8 GHz Pentium 4 PC
with 1 GB memory.
Group 1 simulation studies the effect of a / s ratio by varying it from 0.01 to
1.0 while g = 0 and s is the same for both the background and target. Figure 3.19
gives two sets of absorption coefficient images recovered using the RTE-based
(Figures 3.19a,c) and DE-based (Figures 3.19b,d) methods for small and large ratios.
We also provide the absorption coefficient images recovered using the RTE-based
(Figures 3.19e) and DE-based (Figures 3.19f) methods for a small ratio ( a / s = 0.01)
10 10
0.02 0.02
5 5
0 0.015 0 0.015
–5 –5
0.01 0.01
–10 –10
–10 –5 0 5 10 –10 –5 0 5 10
(a) (b)
10 10
0.02 0.02
5 5
0 0.015 0 0.015
–5 –5
0.01 0.01
–10 –10
–10 –5 0 5 10 –10 –5 0 5 10
(c) (d)
10 10
0.02 0.02
5 5
0 0.015 0 0.015
–5 –5
0.01 0.01
–10 –10
–10 –5 0 5 10 –10 –5 0 5 10
(e) (f )
FIGURE 3.19 Reconstructed absorption images from simulated data using the RTE-based
(a, c) and DE-based (b, d) methods. (a) and (b): μa /μ′s = 0.01. (c) and (d): μa /μ′s = 1.0. (e) and
(f): μa /μ′s = 0.01 with 10.0% noise level.
with 10.0% noise added to the measured data. We see that both methods can recover
the absorption images qualitatively for the two cases, while it is clear that the RTE-
based method has less noise sensitivity than the DE-based method. Quantitative
absorption coefficient profiles along a cut-line (Y = 0 mm) through the target for the
images shown in Figure 3.19 are presented in Figures 3.20a,b. We immediately note
that both methods agree well for a small a / s ratio when the absorption coefficient
is much smaller than the reduced scattering coefficient (Figure 3.20a), while the
absorption coefficient is considerably overestimated using the DE-based method for
a large a / s ratio (Figure 3.20b). We performed several additional reconstructions
0.03 0.03
Recovered Absorption Coefficient

Exact Exact
PAT + RTE PAT + RTE
PAT + DE PAT + DE
0.02 0.02
0.01 0.01
–15 –10 –5 0 5 10 15 –15 –10 –5 0 5 10 15

X Position (mm) X Position (mm)
(a) (b)
50
PAT + RTE
40 PAT + DE
Relative Error (%)
30
20
10
0.00 0.25 0.50 0.75 1.00

Ratio of µa/µs
(c)
FIGURE 3.20 Recovered absorption coefficient profiles along y = 0 mm for the images
shown in Figure 3.19(a): μa /μ′s = 0.01. (b): μa /μ′=
s 1.0. (c): Relative errors of the recovered
absorption coefficient of the target over different μa /μ′s ratios.
with different μa /μ′s ratios, and present the relative errors of the recovered absorption
coefficient of the target for these cases in Figure 3.20c. It can be seen that the error
for the DE-based method increases as the μa /μ′s ratio increases and can be as large as
>40% when μa /μ′s = 1.0. In these cases, the RTE-based method always gives accurate
reconstruction with a largest error of 0.1%.
Group 2 simulation is intended to investigate how the g factor affects the recov-
ered absorption coefficient quantitatively when the μa /μ′s ratio is small (μ′s = 1.0mm–1).
Figures 3.21a and 3.21b present the recovered absorption coefficient images for g =
0.75 using the RTE- and DE-based methods, respectively, while Figure 3.21c shows
the absorption property profiles along the cut-line y = 0 mm for the images shown
in Figures 3.21a and 3.21b. We can see that the DE-based method can qualitatively
reconstruct the absorption image, but significantly underestimate the absorption coef-
ficient values. Relative errors of the recovered absorption coefficient of the target for
several cases with different g factors are given in Figure 3.21d where we see that the
DE-based method underestimates the absorption coefficient values for all the cases,
while the RTE-based method recovers these values accurately. It is interesting to
10 10
0.02 0.02
5 5
0 0.015 0 0.015
–5 –5
0.01 0.01
–10 –10
–10 –5 0 5 10 –10 –5 0 5 10
(a) (b)
Exact 0
PAT + RTE
0.02 PAT + DE
Relative Error (%)
–10
–20
–30
0.01
–40 PAT + RTE
PAT + DE
–50
–15 –10 –5 0 5 10 15 0.00 0.25 0.50 0.75
X Position (mm) g factor
(c) (d)
FIGURE 3.21 Reconstructed absorption images from simulated data with g = 0.75 using
the RTE (a) and DE (b). (c) Absorption coefficient profiles along Y = 0mm for the images
shown in (a) and (b). (d) Relative errors of the recovered absorption coefficient of the target
over different g factors.
note that the error due to the DE-based estimation linearly increases as the g factor
increases.
Group 3 simulation aims to see how the domain size impacts the reconstructions.
In these cases we significantly decrease the domain size from R = 15mm to R = 1.5 mm,
while keeping the ratio R/r as constant, g = 0 and μ′s = 1.0 mm–1. Figures 3.22a and
3.22b show the recovered absorption coefficient images for R = 1.5 mm using the RTE-
and DE-based methods, respectively, while Figure 3.22c gives the absorption property
profiles along the cut-line Y = 0 mm. Figure 3.22d provides the relative errors of the
recovered absorption coefficient of the target with changed domain size. As can be seen
from Figure 3.22, while the DE-based method can reconstruct the absorption images
qualitatively for a small domain size, it always overestimates the recovered absorption
coefficient values for small size domains, and the error can be as large as 35% when the
domain size is just couple of millimeters in diameter. We note that there is almost no
error for reconstructing the absorption coefficient using the RTE-based method.
1 1
0.02 0.02
0.5 0.5
0 0.015 0 0.015
–0.5 –0.5
0.01 0.01
–1 –1
–1 –0.5 0 0.5 1 –1 –0.5 0 0.5 1

(a) (b)
0.030 50
Exact PAT+RTE
PAT+RTE 40 PAT+DE
0.025 PAT+DE
Relative Error (%)
30
0.020
20
0.015
10
0.010 0
–1.5 –1.0 –0.5 0 5 10 15 0 3 6 9 12 15

X Position (mm) R (mm)
(c) (d)
FIGURE 3.22 Reconstructed absorption images from simulated data with R = 1.5 mm
using the RTE (a) and DE (b). (c) Absorption coefficient profiles along Y = 0 mm for the
images shown in (a) and (b). (d) Relative errors of the recovered absorption coefficient of the
target over different domain size.
3.4.3.2.2 Experiments
The PAT setup employed is described in Chapter 1. Here a Ti-Sapphire laser (Altos,
Bozeman, Montana) with a pulse repetition rate of 10 Hz and a pulse width of <10 ns
at 880 nm was used in the experiments. The laser beam was expanded to a beam
with 3.8 cm in diameter and delivered to the top surface of a cylindrical solid phan-
tom. The solid phantoms used in our experiments were made of Intralipid, India
ink, distilled water and Agar power, among which Intralipid served as the scatter-
ing media and the India ink served as the absorption media. In the experiments,
we embedded a single cylindrical target (centered or slightly off-centered) with
various size and optical properties in a cylindrical background having different
size and optical properties. Table 3.1 lists the dimensions and optical properties of
the target and the background used for the five phantom experiments designed for
the comparative study of the RTE- and DE-based qPAT. A mesh consisting of 1301
TABLE 3.1
Optical Properties and Size of the Target and the Background Used in the
Phantom Experiments
Background Targets
Diameter 𝛍a 𝛍′s Diameter 𝛍a 𝛍′s

Cases (mm) (mm–1) (mm–1) g (mm) (mm–1) (mm–1) g
1 6 0.01 1.0 0.59 1.5 0.05 1.0 0.59
2 2 0.01 1.0 0.59 0.5 0.05 1.0 0.59
3 6 0.01 0.05 0.59 1.5 0.05 0.05 0.59
4 2 0.01 0.05 0.59 0.5 0.05 0.05 0.59
5 6 0.01 0.05 0.88 1.5 0.05 0.05 0.88
nodes and 2480 triangular elements along with 64 angular directions was used for
the reconstructions.
Figures 3.23a to 3.23j present the reconstructed absorption coefficient images
for the five phantom experiments using the RTE-based (left column of Figure 3.23)
and DE-based (right column of Figure 3.23) algorithms, respectively. Consistent
with the simulation findings shown in the previous section, we immediately see that
the target can be clearly imaged by the two methods for all the cases. The major
difference between the two methods lies in the ability of quantitatively resolv-
ing absorption coefficient. This difference in quantifying absorption coefficient is
clearly demonstrated in Figure 3.24, which shows the recovered absorption prop-
erty profiles along a transect through the center of the target and background for
the images shown in Figure 3.23. We see that the DE-based method either under-
estimates or overestimates the absorption coefficient of the target for a decreased
small domain size or a large μa /μ′s ratio/increased g factor, while the RTE-based
method always provides accurate reconstruction of the absorption coefficient for
all the cases. Again, these observations are consistent with the findings from the
simulations.
We found that the average value of the recovered absorption coefficient of target
is 0.0511 mm–1, 0.0486 mm–1, 0.0489 mm–1, 0.0486 mm–1, and 0.0492 mm–1 using
the RTE-based method, which is in excellent agreement with the actual value of
0.05 mm–1, while the DE-based method gives a recovered target absorption coef-
ficient of 0.0452 mm–1, 0.0386 mm–1, 0.0565 mm–1, 0.0751 mm–1, and 0.0369 mm–1
for the experimental cases. We list the relative errors of the recovered absorption
coefficient of the target for the five experiments in Table 3.2. It is clear that the
DE-based method presents significant errors for the cases examined especially when
the domain size is decreased and g factor is nonzero (case 2), g factor is increased
(case 5), or the domain size is decreased and the μa /μ′s ratio is increased along with a
nonzero g factor (case 4). We also note that the RTE-based method gives an error of
below 3% for all the cases.
5 5
0.05 0.05
0.04 0.04
0 0.03 0 0.03
0.02 0.02
0.01 0.01
–5 –5
–5 0 5 –5 0 5
(a) (b)
1.5 0.05 1.5 0.05

1 1
0.5 0.04 0.5 0.04
0 0.03 0 0.03
–0.5 0.02 –0.5 0.02
–1 –1
–1.5 0.01 –1.5 0.01
–1 0 1 –1 0 1
(c) (d)
5 5
0.05 0.05
0.04 0.04
0 0.03 0 0.03
0.02 0.02
0.01 0.01
–5 –5
–5 0 5 –5 0 5
(e) (f )
1.5 0.07 1.5 0.07

1 0.06 1 0.06
0.5 0.05 0.5 0.05
0 0.04 0 0.04
–0.5 0.03 –0.5 0.03
–1 0.02 –1 0.02
–1.5 0.01 –1.5 0.01
–1 0 1 –1 0 1
(g) (h)
5 5
0.05 0.05
0.04 0.04
0 0.03 0 0.03
0.02 0.02
0.01 0.01
–5 –5
–5 0 5 –5 0 5
(i) (j)
FIGURE 3.23 Reconstructed absorption coefficient images (mm–1) from the five sets of
phantom experimental data using the RTE (a, c, e, g, i) and DE (b, d, f, h, j): (a) and (b) case 1,
(c) and (d) case 2, (e) and (f) case 3, (g) and (h) case 4, (i) and (j) case 5.
Absorption Coefficient (mm–1) 0.07 0.07

Exact Exact
0.06 PAT+RTE 0.06 PAT+RTE
PAT+DE PAT+DE
0.05 0.05
0.04 0.04
0.03 0.03
0.02 0.02
0.01 0.01
0.00 0.00
–2 –1 0 1 2 –1.5 –1.0 –0.5 0.0 0.5 1.0 1.5
x Position (mm) y Position (mm)
(a) (b)
0.07 0.10

Exact 0.09 Exact

0.06 PAT+RTE PAT+RTE
PAT+DE 0.08 PAT+DE
0.05 0.07
0.04 0.06
0.05
0.03 0.04
0.02 0.03
0.02
0.01
0.01
0.00 0.00
–4 –3 –2 –1 0 1 2 3 4 –1.5 –1.0 –0.5 0.0 0.5 1.0 1.5
y Position (mm) y Position (mm)
(c) (d)
0.07
Exact
0.06 PAT+RTE
PAT+DE
0.05
0.04
0.03
0.02
0.01
0.00
–4 –3 –2 –1 0 1 2 3 4
y Position (mm)
(e)
FIGURE 3.24 Recovered absorption coefficient profiles along (a) x = 0 mm for the images
shown in Figures 3.23(a) and 3.23(b). (b) x = –0.25 mm for the images shown in Figures 3.23(c)
and 3.23(d). (c) x = –0.5 mm for the images shown in Figures 3.23(e) and 3.23(f). (d) x = –1.0 mm
for the images shown in Figures 3.23(g) and 3.23(h). (e) x = 0.0 mm for the images shown in
Figures 3.23(i) and 3.23(j).
TABLE 3.2
Relative Errors of the Recovered Absorption
Coefficient of the Target for the Five
Experimental Cases
Cases DE-Based Algorithm RTE-Based Algorithm
1 –9.6% 2.2%
2 –22.8% –2.8%
3 13.2% –2.2%
4 50.2% –2.8%
5 –26.2% –1.6%
3.5 PARALLEL COMPUTATION
PAT image reconstruction, especially full 3D reconstruction, requires significant
computational times. Given the large problem size, three major areas of the algo-
rithm are costly (see Chapter 1): (1) computation of the forward solution, (2) buildup
of the Jacobian matrix, which requires calculation of the derivatives of acoustic field
with respect to optical/acoustic properties at the receiver sites, and (3) computation
of a full matrix equation for the inverse solution. Because of the massively parallel
structure of these types of image reconstruction algorithms, parallel computing can
be adapted for 3D PAT reconstruction.
Parallel computing is basically the concurrent execution of many computations
using many processors. In a parallel computing scheme, the memory request and
the computation tasks for a computation assignment can be spread on multiple pro-
cessors as shown in Figure 3.25. Here we give a detailed description on the parallel
computing based 3D PAT algorithm. In the frequency-domain FE-based PAT algo-
rithm, the forward calculation of acoustic field and the determination of Jacobian
A computation assignment
CPU CPU CPU CPU CPU CPU CPU CPU
… …
CPU CPU CPU CPU CPU CPU CPU CPU
Memory Memory Memory Memory
FIGURE 3.25 Schematic of parallel computing with a combination of both shared memory
and distributed memory.
matrix J are based on the forward matrix A at each frequency ωi, ordered from ω1
to ωKwith total K frequency elements. The matrix A in the forward calculation is
usually a symmetric sparse matrix and stored in memory by banded storage or com-
pressed storage strategy, requiring less memory than the Jacobian matrix J and the
Hessian matrix JTJ, which is usually a full matrix. Herein, we spread the forward
calculation of acoustic field and the determination of Jacobian matrix on distributed
processors (total number is (Q + 1) by the frequency element ωi, where the whole
matrix A under certain frequency elements is stored in the memory of the specifi-
cally assigned processor. As shown in flowchart of the high-performance photo-
acoustic tomography (Figure 3.26), the matrix A at frequency ωi is stored in the
memory of a processor determined by Mod(i − 1, L), where L is the average number
Assign initial values for Ψ (r)
CPU 0 CPU 1 CPU Q

ωi (i = 1 ~ L) ωi (i L 1 ~ 2L) ωi (i Q L 1~ K )
Solve p by Solve p by Solve p by

[A] { p} = {b} A p b A p b
Solve S si by Solve S si by Solve S si by

δs, j S si Aij δs, j S si Aij δs,j S si Aij
s 1 ~ M, i 1 ~ N s 1 ~ M,i 1 ~ N s 1 ~ M,i 1 ~ N
Fill sub-matrix J0 Fill sub-matrix J1 Fill sub-matrix JQ
Communicate among CPUs and organize sub-matrix Ji (i 0 ~ Q)
Calculate error function e p (c ) (r ) p(m)( r )
Is error small Yes

Stop
enough?
No
Communicate among CPUs and Fill Hessian sub-matrix
Solve the inverse equations in parallel and update Ψ (r)
FIGURE 3.26 Flowchart of the high-performance, parallel FE-based image reconstruction.

of frequency elements on each processor (total frequency elements K divided by

total processors Q + 1). In each processor, the forward modeled acoustic field as well
as the elements of Jacobian matrix at the frequencies associated with the specified
processor is calculated independently, after which the Jacobian matrix J over the
whole frequency range is assembled and stored dispersedly in the memories of the
distributed processors. In this way of parallel computing and storage, the computa-
tion load and storage is evenly assigned among the processors, and the computation
assignment is maximally parallelized since each processor can run the assigned task
independently without communicating with other processors.
The Jacobian matrix J and the Hessian matrix JTJ (denoted as H) are full matrices,
requiring distributed storage over the processors. A wrapping storage over the col-
umns of the Jacobian matrix J is used to evenly store the Jacobian submatrix over the
processors and further efficiently calculate the Hessian matrix H with minimal mutual
communication between the processors. The wrapping storage is described below,

{J j} ∈CPU Myid , if Myid = Mod (j − 1, Q + 1) (3.30)

where {J j } = {J1 j ,…, J Mj , J( M +1) j ,…, J( M × K ) j }T , j = 1,…, N .
With the wrapping storage of the Jacobian matrix J, the Hessian matrix H can be
calculated and stored by

{H j} = {J1 J2 J j}T {J j} ∈CPU Myid , if Myid = Mod (j − 1, Q + 1) (3.31)

where {H j } = {H1 j ,…, H jj }T , j = 1,…, N .
Because the Hessian matrix H is symmetric, we store only the upper triangle
of the Hessian matrix spread over the processors. In the calculation of element Hij,
communication and data exchange
may
be involved among the distributed proces-
sors, except that the vector {J j } and {J j } are both stored in the memory of the same
processor. Since the Hessian matrix is stored dispersedly, the inverse reconstruction
requires a parallel solving method, where parallelized Cholesky decomposition is
used in the high-performance PAT. While the parallelized code can improve the
computation speed at least by two-order of magnitude compared to the un-paral-
lelized code for 3D image reconstruction, it is still computational demanding. For
example, a 3D image reconstructed using a finite element mesh of 41,323 tetrahedral
elements and 7519 nodes required ~120 minutes per iteration for a total of 10 itera-
tions on a computation cluster with 10 processors (each of the processors worked at
2.2 GHz with 2 GB memory).
3.6 VARIABLE-THICKNESS MULTILAYERED
POLYVINYLIDENE FLUORIDE TRANSDUCER
In PAT, the ultrasound transducer is a critical component as it receives the wide-
band acoustic waves generated in tissue by the pulsed laser. The sensitivity of the
transducer determines largely the signal-to-noise ratio (SNR), while its bandwidth
defines the resolution. Transducers currently used for PAT are largely based on
three types of piezoelectric materials including piezoelectric ceramics (PZT), sin-
gle crystal (PMN-PT), and polyvinylidene fluoride (PVDF). Among these materi-
als, PVDF owns some advantages. PVDF has a high mechanical damping ability
and a unique permittivity, resulting in a high bandwidth. In addition, the acoustic
impedance (Z0) of PVDF is 2.7 MRays, which is much lower than that of PZT
(>30 MRays). Hence it provides a good coupling efficiency to human tissue or com-
monly used ultrasound gels. PVDF is flexible and can be easily fabricated in convex
shapes replacing the acoustic lens commonly used in PZT or PMN-PT. We note,
however, that the sensitivity of the PVDF transducers currently used for PAT is
not comparable to the PZT and PMN-PT based transducers (Manohar et al. 2004,
Ermilov et al. 2009, Xi et al. 2012), leading to limited SNR especially when targets
are deeply located within tissues. In order to improve the sensitivity of the PVDF
transducers, several transducer designs based on the same-thickness PVDF layers
including folded transducer (FT), Barker-coded transducer (BCT), and switchable
Barker-coded transducer (SBCT) have been suggested in the field of ultrasound
imaging (Zhang and Lewin 1995, Fukuda et al. 2006, Nakazawa et al. 2007). While
the sensitivity is improved, there is no significant improvement in bandwidth from
these designs. Here we describe a design of PVDF transducer with variable-thick-
ness layers to provide improvement in both the sensitivity and bandwidth for pho-
toacoustic imaging.
The design of our multilayered transducer is shown in Figure 3.27a where the
PVDF films (metalized film sheets, measurement specialties) having three different
thicknesses (28 µm, 52 µm, and 110 µm) are sandwiched by electrodes. The fabrica-
tion procedure of this transducer is as follows: (1) shaping the lateral dimension of
Electrode –
PVDF (28 µm) Trigger Mirror
+ Nd: YAG
+ Transparent glue
PVDF (52 µm)
–
– Rotator
PVDF (110 µm) 2D stage
+
PC
Backing
(a) Pulser/receiver
Lens
Amplifier
Trigger Transducer
Transducer
Oscilloscope (c)
(b)
FIGURE 3.27 (a) Configuration of the multilayered PVDF transducer. (b) Schematic of
the circular-scanning based photoacoutic imaging system. (c) Schematic of the pulse/echo
evaluation system.
the three PVDF films to 10 mm × 3 mm; (2) removing the unneeded silver coating
from both sides of each film to ensure electric insulation between the positive and
negative electrodes; (3) stacking/gluing the films/backing material together using
ultrasound transparent epoxy (EPO-TEK 301-2, Epoxy Technology); (4) connecting
coaxial cables to the electrodes using silver epoxy (E-solder 3022. VonRoll USA,
Inc.); and (5) sealing the whole transducer with ultrasound transparent epoxy.
The multilayered transducer fabricated is tested using a circular scanning pho-
toacoustic imaging system (Figure 3.27b). A laser beam generated from a pulsed
Nd:YAG laser (532 nm) was redirected through the center of the rotator and
expanded by a lens to illuminate an imaging area of 6.25 cm2. A 2D linear stage
(NLS4 Series, Newmark Systems) was used to optimize the position of transducer.
A water-tank with a through hole in the bottom sealed with transparent membrane
was used to provide coupling for ultrasound transmission. The signal received by
the PVDF transducer was amplified by two low-noise amplifiers and digitalized/
stored by a data acquisition card at 50MHz sampling rate. A pulse/echo test system
(Figure 3.27c) was used to evaluate the bandwidth of the multilayered transducer.
The transducer was driven by a pulser/receiver (5073PR, Panametrics-NDT), and a
piece of metal with a thickness of 5 mm was placed in front of the transducer to act
as a reflector. The reflected pulse was received by the transducer, amplified by the
internal amplifier of the pulser/receiver and recorded using an oscilloscope.
Before the performance evaluation and in vivo experiments, the transducer was
calibrated through an experiment with a hair-embedded solid phantom. The solid
phantom was a mixture of Intralipid (scatter), India ink (absorber), and 2% Agar
powder (solidifier). The absorption and reduced scattering coefficients of the phan-
tom were 0.007 mm–1 and 1.0 mm–1. A human hair was inserted in the solid phantom,
which was illuminated by the light beam. Figures 3.28a, 3.28b, and 3.28c show the
acoustic signals from different layers of the transducer where we note a time shift
1000 1000
28 um T1 = 0.6 52 um T2= 0.66
PA Signal
PA Signal
0 0
–1000 –1000
0 0.4 0.8 1.2 0 0.4 0.8 1.2
Time (us) Time (us)
(a) (b)
1000 1000 W Cali

110 um T3 = 0.7
PA Signal
PA Signal
W/o Cali
0 0
–1000 –1000
0 0.4 0.8 1.2 0 0.4 0.8 1.2
Time (us) Time (us)
(c) (d)
FIGURE 3.28 Photoacoustic signals of the 28µm- (a), 52 µm- (b), and 110 µm-thick (c) layer
transducer from a hair. (d) Photoacoustic signals with and without the calibration.
between these signals, caused by the different distance of these three layers from the
target. Hence, calibration of this time shift is needed before further evaluating the
performance of the transducer. Assuming that the arrival times of the signals are T1,
T2, and T3 for the layers of 28, 52, and 110 µm, respectively, and setting the arrival
time for the 28 µm-thick layer as the baseline, we can then calculate the time shift of
the other two layers relative to T1 by the following relationships:
Δt1 = T2 − T1 (3.32a)
Δt2 = T3 − T1 (3.32b)
Thus, the photoacoustic (PA) signal from the multilayered transducer can be
obtained as
PA(t) = PA28μm(t) × P1 + PA52μm(t − Δt1) × P2 + PA110μm(t − Δt2) × P3 (3.33)
where P1, P2, and P3 are the phase of the PA signals which are 1, –1, and 1 in
our case for the 28 µm-, 53 µm-, and 110 µm-thick layer, respectively. Figure 3.28d
shows the calibrated signal and noncalibrated signal. We see that the SNR of the
calibrated signal is improved significantly (by 2.5 times) compared with that without
the calibration.
Figures 3.29b, 3.29c, and 3.29d show the frequency characteristics by fast Fourier
transforming the reflected acoustic waves detected by each of the three layers using
the pulse/echo system (Figure 3.27c). For the 110 µm-, 52 µm-, and 28 µm-thick
single-layer transducer, the central frequency and the bandwidth (-– level) are,
respectively, 5 MHz/85%, 10 MHz/78%, and 15 MHz/59%, while the central fre-
quency of the multilayered transducer is 10 MHz and the –6 dB level bandwidth is
improved to be 140%, as shown in Figure 3.29a.
In vivo animal experiments were performed to evaluate the merits of the improved
bandwidth and sensitivity for photoacoustic imaging. The brain of adult BALB/C mice
(~25g) was imaged. Before imaging, the hair was removed with a hair remover lotion.
The mice were anesthetized by a mixture of ketamine (85 mg/kg) and xylazine during
the imaging, and were sacrificed afterward using the University of Florida Institutional
Animal Care and Use Committee (IACUC)-approved techniques. The laser beam pro-
vided an incident energy density of 9 mJ/cm2 at the skin of the mice, which is much
lower than the safety standard provided by ANSI (20 mJ/cm2 at 532 nm). For each
imaging experiment, we collected PA signals at 120 positions by scanning the trans-
ducer around the mouse brain with a scanning step of 3°. The PA image was recon-
structed by the delay-and-sum algorithm described in Chapter 1.
Figure 3.30a shows a typical signal from in vivo mouse brain measured by the
multilayered transducer without (top panel) and with (bottom panel) calibration.
Without the calibration, the SNR is low and the image formed with such signals has
relatively poor quality. As we see from Figure 3.30b (left), in such case, some blood
vessels inside the brain are not visualized and almost every blood vessel imaged
is not continuous, as compared to the photograph of the brain after removal of the
scalp (Figure 3.30b, right). With the calibration, the SNR is significantly improved as
Frequency Response (Multilayer) Frequency Response (110 µm)

1
1
0.9
Bandwidth Bandwidth
0.8 0.8 (80%)
(140%)
0.7
0.6 9 MHz
0.6 18 MHz
0.5 0.4
0.4
0.2
0.3
4 MHz
0.2 0 4 MHz
0 10 20 30 40 0 10 20 30 40
Frequency (MHz) Frequency (MHz)
(a) (b)
Frequency Response (52 µm) Frequency Response (28 µm)
1 1
Bandwidth Bandwidth
0.8 (78%) 0.8 (59%)
0.6 0.6
13.8 MHz
0.4
0.4
0.2
0.2
0 10 MHz 18.5 MHz
6 MHz
0
0 10 20 30 40 0 10 20 30 40
Frequency (MHz) Frequency (MHz)
(c) (d)
FIGURE 3.29 Frequency response of the multilayered PVDF transducer (a); 110 µm-thick
layer PVDF transducer (b); 52 µm-thick layer PVDF transducer (c); and 28 µm-thick layer
PVDF transducer (d).
shown in the bottom panel of Figure 3.30a. From the image formed with the calibra-
tion (Figure 3.30b, middle), we see that the eyes, blood vessels and other tissues are
identified clearly in comparison with the photograph.
In a separate in vivo experiment, we compared the sensitivity of the multilayered
transducer with the single-layer transducers, and the PA images obtained are pre-
sented in Figure 3.31. We immediately note the clearly higher SNR shown by the
image using the multilayered transducer (Figure 3.31a) compared with that using the
single-layer transducer (Figures 3.31b, 3.31c, and 3.31d). In particular, we can iden-
tify the tissues, organs, and blood vessels inside the brain from Figure 3.31a much
more clearly than that from Figure 3.31b, 3.31c, and 3.31d as indicated by red arrows.
We have shown that the design of variable-thickness multilayered PVDF trans-
ducer provides significant improvement in both the sensitivity and bandwidth com-
pared to the conventional single-layer transducer or commercial hydrophone (e.g.,
from Precision Acoustics, England). However, we note several limitations associated
with the current design, which should be overcome in the future. First, the sensitivity
Multilayer without Calibration

1000
500
0
–500
–1000
0 500 1000 1500 2000 2500 3000
Multilayer with Calibration

1000
500
0
–500
–1000
0 500 1000 1500 2000 2500 3000
(a)
Eye
Max Eye
Min
(b)
FIGURE 3.30 (a) In vivo photoacoustic (PA) signals from the multilayered transducer with
(bottom panel) and without (top panel) calibration. (b) Reconstructed brain images using PA
signals without (left panel) and with (middle panel) calibration. Photograph of the brain after
the removal of the scalp (right panel).
of the current multilayered transducer is still not comparable with that of a commer-
cial PZT unit. This, however, can be overcome by stacking more layers (more than
three) together to make a more sensitive transducer. Second, the PVDF films used
in the current work are commercial units with electrodes on both sides of each film.
After stacking them together, there are two layers of electrodes between two adjacent
films. As a result, the improvement of the sensitivity is not linearly proportional to
the number of layers stacked due to the attenuation from the electrodes. This prob-
lem can be solved by using raw film liquid, which will allow us to use only one much
thinner electrode (just a few µm) between two adjacent films during depositing the
films. Third, no matching layer and electromagnetic shield (EMS) housing were used
in the current design. Hence, we had to average 50 times on the signals collected to
reduce the electromagnetic noise, resulting in a relatively long data acquisition time.
Multi with Calibration 110 um

Max
Min
(a) (b)
52 um 28 um
(c) (d)
FIGURE 3.31 In vivo photoacoustic imaging of mouse brain using the multilayered trans-
ducer (a) and single-layer transducer (b: 110 µm thick; c: 52 µm thick; d: 28 µm thick).
3.7 ALN-BASED PIEZOELECTRIC MICROMACHINED

ULTRASONIC TRANSDUCER
Capacitance micromachined ultrasonic transducer (cMUT) has been used for PAT
(Vaithilingam et al. 2006). However, the capacitance of cMUT is low, and its per-
formance is susceptible to the influence of parasitic capacitance, especially when it
is integrated with system-on-chip (SOC). Piezoelectric micromachined ultrasonic
transducer (pMUT), in comparison, has higher capacitance, which implies that the
effect of a system parasitic capacitance on the coupling and sensitivity is not as
significant as in the case of cMUT. Although pMUTs have lower electromechani-
cal coupling, they do not need the high polarization voltages and small capacitive
gaps required by cMUTs (Guldiken et al. 2009, Huang et al. 2009). Studies in lead
zirconate titanate (PZT)–based pMUT were reported recently and used for pure
ultrasound imaging (Dausch et al. 2008, Wu et al. 2008). However, the PZT pro-
duction generally requires high temperature that is not compatible with integrated
circuit (IC). In comparison with PZT, the lower piezoelectric coefficients of AlN are
mitigated by a significantly reduced dielectric constant (Muralt 2000). The reduced
capacitance of AlN-based pMUT can provide improved SNR.
Polyimide
film
Upper
electrode
Lower AlN film
electrode
SiO2
FIGURE 3.32 Structural schematic of AlN-based pMUT.
Here we describe the fabrication and characterization of the pMUT and its appli-
cation to PAT. As shown in Figure 3.32, the pMUT consists of SiO2, bottom elec-
trode, AlN films, upper electrode, and polyimide (PI). In the microstructures, PI
was fabricated using spin-coating and reactive sputtering technology. PI played two
important roles here: one was to isolate the upper and lower electrodes and to reduce
the possible parasitic capacitance between the two electrodes; the other was to act
as the flexible protective layer on the surface. Compared to SiO2 protective layer, PI
is conducive to the vibration of the films according to the finite element simulation.
In the design of pMUT, AlN-based piezoelectric layer was synthesized through
middle-frequency magnetron reactive sputtering from metallic rectangular Al tar-
gets at room temperature compatible with the standard IC technology. Figure 3.33a
shows a typical field emission scanning electron microscopy (SEM) image of the
cross-sectional structure of the AlN films on the bottom electrode. The SEM image
reveals that the AlN thin film was well-crystallized and comprised well-oriented
columnar grains that were perpendicular to the substrate. The diameter of the colum-
nar grains was approximately 140 nm. Figure 3.33b is the 2θ-ω scan x-ray diffrac-
tion (XRD) spectrum of the AlN thin films, which shows only the (0002) hexagonal
AlN and Pt (111) peaks. The inset shows the rocking curve of the (0002) diffraction
peak for the AlN thin films. The FWHM of the rocking curve is 7.0°.
To increase the receiving response, the pMUT had an area ratio of 0.45 between
the top electrode and vibrating membrane, and the AlN film had only a 300 nm-
thick silicon oxide film and a 200 nm-thick bottom electrode as its supporting layer.
Our experience showed that acrylonitrile-butadiene-styrenecopolymer (ABS) could
protect the superficial silicon oxide in tetramethylammonium hydroxide (TMAH)
solution for about 8 hours when the annealing temperature was 250°C. The rela-
tionship between the protection time and temperature of etching solution is given
in Figure 3.34. Based on this experience, we constructed the pMUT with only 300
nm-thick silicon oxide layer as the device supporting layer.
The electromechanical coupling coefficient of pMUT is defined as (ANSI/IEEE
Std 176-1987):
2
keff fa2 − fr2
2 = (3.34)
1 − keff fr2

1 µm
(a)
X-ray Diffraction Intensity (a.u.)
AlN (0002)
X-ray Diffraction Intensity (a.u.)
Pt (111)
5 10 15 20 25 30
Omega (deg)
20 30 40 50 60
2 Theta-Omega (deg)
(b)
FIGURE 3.33 (a) SEM image of the AlN on bottom electrode. (b) 2θ-ω scan X-ray diffrac-
tion (XRD) spectrum of the AlN thin films.
where fr is the resonance frequency and fa the antiresonance frequency in the imped-
ance frequency spectrum. As shown in Figure 3.35, the impedance and phase frequency
spectrum of the pMUT was characterized using an impedance analyzer (Agilent 4294,
Agilent Technologies, Inc), fa and fr were measured to be 2.885 MHz and 2.93 MHz,
respectively, with 0.005 MHz frequency sweep step. The calculated coupling coefficient
is 2.38% ~ 3.71%, compared to a value of 0.056% for AlN-based pMUT which has a
high parasitic capacitance (Shelton et al. 2009). However, it is generally lower than PZT-
based pMUT (Akasheh et al. 2004, Muralt et al. 2005). In addition, the small phase
change seen from Figure 3.35 means a small value of ke2Q , which might be resulted
from the parasitic capacitance between the lower electrode and upper electrode lead.
12
Protection Time of ABS/h

10
4
70 75 80 85 90
Temperature of TMAH Solution (25%)
FIGURE 3.34 The relationship curve of protection time and temperature of TMAH.
To characterize the signal detection capability of the pMUT, we configured it to

detect photoacoustic signals. Figure 3.36 gives the schematic of the photoacoustic
imaging system with which the pMUT was integrated. An Nd:YAG 532 nm laser
(Brilliant, QUANTEL) was employed to generate pulses of 7ns duration at 5Hz. The
focused beam was coupled into a 62.5 µm multimode fiber and focused again by a self-
focusing lens at the output. The laser spot size was 220 µm and the laser energy was
below 10 mJ/cm2. The ultrasonic couplant was used as the ultrasonic matching layer
between the pMUT and phantom to be conducive to the signal detection. The resolu-
tion of this photoacoustic imaging system is related to the spot size of the focused laser
beam used—a more tightly focused laser beam provides a higher resolution.
32400 87.6
31200 88.0
Impedance (ohm)
Phase (degree)
30000 88.4
28800 88.8
27600 89.2
2.6 M 2.8 M 3.0 M 3.2 M 3.4 M

Frequency
FIGURE 3.35 Impedance- and phase-frequency curves.

Coupler
Nd: YAG
Focusing lens
Optical fiber
PC
2D translation stage
Self-focusing lens
Phantoms
DAQ card pMUT
Amplifier
FIGURE 3.36 Schematic of the pMUT-based photoacoustic imaging system.
The self-focusing lens was scanned for 80 steps with a step size of 50µm along
x and y axes. In the experiment, a piece of human hair was embedded in a stan-
dard agar-powder solidified tissue-mimicking phantom, consisting of Intralipid as
the scatterer and India ink as the absorber. The photoacoustic signals transmitted
through the phantom and coupling agent and reached to the device. Figure 3.37a
shows the photoacoustic signal from the human hair, while Figures 3.37b and 3.37c
present the photograph and reconstructed photoacoustic image of the phantom. We
see that the signal from the hair is strong and the image is well recovered. Compared
with other systems, pMUT has a significant impact on photoacoustic imaging
because it has wide operating bandwidth, can be configured as large array, and can
be easily integrated with electronics.
The pixel amplitude profile of the recovered hair (Figure 3.37c) is shown in
Figure 3.38. This allows us to calculate the average FWHM of the profile or the lateral
resolution of the system to be 240 µm. It is, however, possible to obtain higher resolu-
tion by using more tightly focused laser beam with the trade-off of penetration depth.
We have shown the design and fabrication of AlN-based pMUT with a high cou-
pling efficiency, and demonstrated the application of such pMUT for photoacoustic
imaging. We believe significant applications of pMUT to PAT will be quickly identi-
fied, especially when the array of pMUTs and their integration with SOC become
available. While the results presented here are encouraging, we note several limita-
tions associated with the current design, which should be overcome in the future.
First, the sensitivity of the current pMUT is still not comparable with that of a PZT
unit. To improve response of AlN-based pMUT, we can increase the piezoelec-
tric coefficient of thin film by doping scandium inside AlN film (Akiyama et al.
2009). Previous studies showed that the transverse piezoelectric coefficient e31, f of
ScxAl1-xN would be augmented with the increased Sc content (Matloub et al. 2013).
Second, the bandwidth of the pMUT needs to be improved for in vivo photoacoustic
imaging. This can be realized by integrating together several sets of elements with
different central frequencies into one chip.
0.2
0
Amplitude (mV)
–0.2
–0.4
–0.6
–0.8
0 500 1000 1500 2000 2500 3000
(a)
1 mm
(b) (c)
FIGURE 3.37 (a) Photoacoustic signal from the hair detected by the pMUT. (b) Photograph
of human hair embedded in the phantom. (c) Recovered photoacoustic image.
1
0.9
0.8
0.7
0.6
Intensity
0.5
0.4
0.3
0.2
0.1
0 10 20 30 40 50 60 70 80
Distance Along Profile
FIGURE 3.38 Intensity profile for the recovered photoacoustic image shown in Figure
3.37(c).
3.8 LIQUID ACOUSTIC LENS

Aiverging (negative) acoustic lens shows a great capability to improve the perfor-
mance of PAT by enlarging the acceptance angle. In this implementation, an acrylic
plastic lens is glued on a flat transducer so that the directivity of the transducer is
modified to cover a larger field of view (FOV). By adding the lens, the PAT system
can reconstruct the image of a larger area with better accuracy (Xia et al. 2011).
Acoustic lens may play an equivalent role in ultrasound emission or detection as
optical lenses in optical system. However, light does not suffer from severe atten-
uation and reflection in most glass-based optical lenses whereas the transmission
efficiency of ultrasound in solid-based acoustic lens is very low owing to the high
acoustic impedance of solid. Antireflection coating on the surface of solid-based
lens could increase the transmission (Niederhauser et al. 2004), but the fabrication
could be time-consuming and not cost-effective. Recently the field of optofluidics
has emerged with the enlightening idea that solid-based optical components could
be replaced by liquids or fluids with better performance and more functionalities
(Psaltis et al. 2006). Since the acoustic impedance of liquid is much lower compared
with solid, building an acoustic lens with liquid can be advantageous. Here we pres-
ent the design and fabrication of a novel liquid acoustic diverging lens for PAT. A
hydrophone is used to map the pressure distribution of a transducer with and without
the proposed liquid acoustic lens. Further experiments using the lens are carried out
to demonstrate the advantages of such a lens in a photoacoustic tomography system.
Figure 3.39 shows the fabrication process of the liquid acoustic lens. A lens mold
was physically built by a 3D printer (Objet Eden 260V) from a predefined 3D digital
file (Figure 3.39a). Polydimethylsiloxane (PDMS) was mixed from the two compo-
nents (silicone elastomer base and its curing agent) with a weight ratio of 10:1. After
the mixing and degassing, PDMS was poured into the mold and heated up to 80°C
for about 40 min (Figure 3.39b). When PDMS was cured after heating, it was peeled
off from the mold and bonded with another piece of PDMS membrane (Figure 3.39c).
The aperture of the lens is determined by the outer diameter of the transducer, which
is 9.2 mm in this work. Liquid can be infused into the lens chamber by a syringe
pump (KDS 210, KD Scientific) through tubing and needle (Figure 3.39d). The liquid
lens as well as the transducer can be mounted on a holder (3D printer fabricated)
with ultrasound gel in between (Figure 3.39e). Basically the lens is formed by a pneu-
matic way, which has been reported for optical applications (Zhang et al. 2004) or
micropump on a chip (Tan et al. 2010). Here we purposely chose glycerol (75%) and
water (25%) mixture as the lensing liquid to apply this configuration for an acoustic
application. The mixture has a density of 1.19g/cm3 and the speed of sound in this
mixture is 1.78mm/µs. The acoustic impedance of this mixture is 2.11Mrayl, which
is lower compared to the impedance of aluminum 17.3Mrayl (Niederhauser et al.
2004) and acrylic 3.23Mrayl (Xia et al. 2011).
In photoacoustic tomography, the acceptance angle of the transducer is a key
factor, which decides the accuracy of the image reconstruction. An ideal transducer
for tomography should have a 360° acceptance angle to cover the whole field of
view. But usually, the directivity of a commercial flat transducer limits the accep-
tance angle. As a result of using a flat transducer, the reconstructed image may lose
PDMS
Hard resin
(a) Lens mold (b) Filling and curing PDMS in the mold
Needle Liquid
(c) Membrane bonded with lens (d) Liquid filled into the chamber
chamber and two holes punched through the hole with needle
through from sides
Holder
Transducer
(e) A holder designed to mount the

lens as well as the transducer
FIGURE 3.39 Fabrication process. (a) A lens mold is built by a 3D printer; (b) PDMS is
poured into the mold and cured with 80°C; (c) a piece of membrane (PDMS) is bonded and
seals the lens chamber; (d) liquid is infused into the chamber through the holes to shape the
curved lens interface; (e) a holder, which is also made by a 3D printer, mounts the lens as well
as the transducer.
the true features of the targets, especially when the targets are located far from the
center of the scanning area. Efforts have been made to enlarge the acceptance angle
by gluing a diverging (negative) acrylic lens with a flat transducer (Xia et al. 2011).
However, the problems associated with this method arise: (1) the transducer can
hardly be reused for other purposes because the removal of the acrylic lens may
damage the transducer; (2) the solid-based lens inherently has a high acoustic imped-
ance, which leads to considerable loss of acoustic energy by reflection.
These problems motivated us to propose and test a liquid acoustic lens. The test-
ing of the lens was implemented in three steps: (1) the transducer worked in its trans-
mission mode and its pressure distributions with and without the liquid lens were
mapped and compared; (2) pencil lead, as an ideal target, was used in photoacoustic
tomography to look into how much image deformation would be caused due to the
limited directivity for both cases with and without lens; (3) phantom was used to
mimic tissue and imaged by PAT for both cases with and without lens. In this test,
the transducer (2.25 MHz, V133, Olympus) was in its transmission mode emitting
ultrasound pulses. A holder mounted the transducer as well as the liquid acoustic
lens and fixed their positions. A hydrophone (HNR-0500 ONDA) was employed to
scan laterally and axially while detecting the ultrasound pulses from the transducer
(a)
Hydrophone
Scanning area
Transducer
Holder Water tank
1
0.9
0.8
0.7
0.6
Without lens
0.5
0.4
0.3
0.2
0.1
With lens
(b) (c)
FIGURE 3.40 (a) Schematic setup for testing the performance of the liquid lens; (b) acous-
tic pressure distribution of the transducer with and without the liquid lens; (c) photograph of
the prototype of the liquid lens mounted on the holder.
(Figure 3.40a). The signal was amplified and acquired by a computer. Figure 3.40b

shows the pressure distribution of the transducer with and without the liquid lens.
The peak-to-peak of the acoustic pressure is represented by the intensity value of
each pixel. And all the values are normalized by the maximum value of the entire
field. Without the lens, the bare transducer produced an acoustic beam, covering
relatively smaller area of the field and being slightly focused. On the other hand,
the acoustic beam was expanded and covered a larger area by attaching the liquid
acoustic lens to the transducer (Figure 3.40b).
Seven pencil leads were aligned in a line and embedded in a background with the
right-most one located at the center of the scanning area (Figures 3.41a and 3.41b).
The background had an absorption coefficient of µa = 0.007 mm–1 and a scatter-
ing coefficient of µs = 1 mm–1. The absorption was introduced by adding ink. The
experimental setup was the same as that described in Section 3.6, except that the liq-
uid acoustic lens was attached to the transducer in this experiment. Delay-and-sum
algorithm was used to reconstruct the images (see Chapter 1).
Figures 3.41a and 3.41b show the comparison between the reconstructed images with-
out and with the lens. In the absence of the lens, the image of the target started to suffer
–10
–5 8
0 4
5 0
10
0 5 10 15 20 25 30
(mm)
(a)
–10
10
–5
0 5
5 0
10
0 5 10 15 20 25 30
(mm)
(b)
Tangential Profile of the Target (Black Arrow)
2.5 With lens

Without lens
2
1.5
Intensity Profile
0.5
–0.5
–4 –3 –2 –1 0 1 2 3 4
(mm)
(c)
FIGURE 3.41 PAT of pencil leads. Seven pencil leads (0.7 mm diameter) are aligned along
the radial direction with the right-most one located at the center of the scanning. (a) Imaging
without liquid lens; (b) imaging with liquid lens; (c) intensity profile along the tangential
direction of the pencil lead with black arrow pointed.
from tangential deformation when its distance from the center was about 5 mm. This
effect of tangential deformation became increasingly magnified to the targets located
further from the center. The other observation was that the image contrast dropped with
the increasing distance of the target from the center. The tangential deformation is ubiq-
uitously presented in photoacoustic imaging because the transducer is not an ideal point
detector. Considering the circularly scanning detection method in most PAT systems,
the effect of tangential deformation could be suppressed as long as the target could be
always covered by the FOV (field-of-view) of the detector. However, in reality, only the
target close to the center could be always covered by the scanning detector. Therefore,
the further the target from the center, the worse its image suffered from deformation.
Enlarging the acceptance angle using a diverging lens is an efficient way to sup-
press the tangential deformation. Figure 3.41b shows that the image of the target
(with black arrow pointed), located at about 15 mm away from the center, remained
an approximately similar shape as its true size. On the other hand, the corresponding
image generated without the lens (Figure 3.41a) went through a severe deformation.
The tangential profiles of the images (with and without the lens) were plotted in
Figure 3.41c. The comparison indicates that the diverging liquid acoustic lens can
efficiently suppress the deformation by enlarging the acceptance angle of the detector.
In the phantom experiments, three targets of phantom, which had an µa =
0.049 mm–1 and a µs = 0.1 mm–1, were positioned at the center (No. 1), with a hori-
zontal distance of 11~mm from the center (No. 2), and with a vertical distance of
10 mm from the scanning plane (No. 3), respectively (Figure 3.42). All the targets
had the same diameter of 4 mm. Figures 3.42a and 3.42b show the reconstructed
images of PAT without and with liquid acoustic lens, respectively. Generally, the
image produced with lens (Figure 3.42b) has a better SNR than that produced with-
out lens (Figure 3.42a). This is because the lens can enlarge the acceptance angle and
cover a larger FOV, and thus the signals from the off-center targets could contribute
more in the linear delay-and-sum reconstruction process.
No. 2 target, located 11 mm away from the center, was imaged to have tangen-
tial deformation without the lens. The comparison between the tangential profiles
in Figure 3.42c could apparently illustrate the deformation. On the other hand, the
reconstruction of image over radial direction did not present any deformation under
either condition with or without the lens (Figure 3.42e). Another interesting inspec-
tion is the imaging of the target (No. 3) off the scanning plane. The image produced
without using the lens presented only a fragment of the target, which is the part close
to the central axis (Figure 3.42a). This is because the bare transducer with a lim-
ited directivity can cover only the signal emitted from the part of the target, which
is close to the central axis of the scanning plane, and the signal of the other part,
which is out of the detection cone of the transducer, can hardly be detected at any
scanning points. This signal-blind-zone resulted in a phenomenon that is different
from the tangential deformation of No. 2 target. On the other hand, by enlarging the
acceptance angle using the lens, the detection cone of the transducer can cover the
entire body of No. 3 target. As a result, the image could be correctly reconstructed
(Figure 3.42b). Figures 3.42d and 3.42f show that the image produced with the lens
remains approximately the true size of 4 mm diameter (tangential and radial), but
the image produced without using lens loses its fidelity by shrinking tangentially and
–20 –20
6 8
No.1 No. 1
–10 –10 6
No. 2 No.3 4 No. 2 No. 3
4
(mm)
(mm)
0 2 0
2
10 0 10 0
–2 –2
20 20
–20 –10 0 10 20 –20 –10 0 10 20
(mm) (mm)
(a) (b)
2
1.8 No. 2 target No. 3 target
Tangential Profile
Tangential Profile
1
1.4
0
1
With lens –1 With lens
0.6 Without lens Without lens
0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7
(mm) (mm)
(c) (d)
2
No. 2 target No. 3 target
1
1
Radial Profile
Radial Profile
0
0
With lens With lens
–1 Without lens –1 Without lens
0 1 2 3 4 5 6 7 0 1 2 3 4 5 6
(mm) (mm)
(e) (f )
FIGURE 3.42 Results of phantom experiments. A target (No. 1) is placed close to the center
for reference. No. 2 target is located in the same plane of No. 1 but with a distance of 11 mm
from the center. No. 3 target is lifted up to have a vertical distance of 10mm from the center.
All the targets have the same diameter of 4 mm. (a) Imaging without lens; (b) imaging with
lens; (c) tangential profile of No. 2 target; (d) tangential profile of No. 3 target; (e) radial pro-
file of No. 2 target; (f) radial profile of No. 3 target.
distorting radially. Therefore, enlarging the detection cone using the liquid acoustic
lens could also be promising to improve the SNR and correct imaging deformation
in 3D photoacoustic tomography (Wang et al. 2012, Xiang et al. 2013), which has
been demonstrated as an effective imaging method for the diagnosis of breast cancer
(Xi et al. 2012, Manohar et al. 2004, Ermilov et al. 2009).
In sum, compared with previous solid-based acoustic lens, the liquid lens presents
inherently low acoustic impedance and the convenience to be attached to or detached
from any commercial transducers without damaging the transducers. The material
PDMS that encapsulates the liquid is biocompatible, which makes this lens suitable
for biomedical applications.
3.9 MICROELECTROMECHANICAL SYSTEMS SCANNING MIRROR

Photoacoustic imaging using miniaturized scanning systems/probes represents a
new direction in this field. Such a miniaturized system allows for high-resolution/
high-contrast photoacoustic imaging in applications such as endoscopic screening
and image-guided surgery. The existing miniaturized probe based photoacoustic
imaging is based on mechanical scanning and is generally a slow approach, limit-
ing it to 2D imaging. On the other hand, microelectromechanical systems (MEMS)
mirrors can manipulate laser beams in free space, eliminating optical coupling non-
uniformity and instability. Two-dimensional MEMS mirrors can also perform fast
roster scanning (Xie et al. 2003, Zara et al. 2003).
In this section, we describe a 2D MEMS-based scanning system for 3D photoacous-
tic imaging. In this system, we use a 2D MEMS mirror based on thermal bimorph
actuation mechanism to realize fast light scanning and a single element ring-shaped
PVDF transducer to detect ultrasound signals. Three-dimensional images of pencil
lead(s) embedded in phantom materials and chicken legs and images of in vivo human
blood vessels are successfully obtained using this MEMS-based photoacoustic system.
3.9.1 Photoacoustic Imaging System

The schematic of the experimental setup is shown in Figure 3.43a, in which a
MEMS mirror (see Figure 3.43d), embedded in a small-diameter scanning probe
(Figure 3.43b), is used to perform laser beam scanning and an Nd:YAG 532 nm
laser (NL 303HT from EKSPLA, Lithuania) is employed to generate pulses of 20 ns
duration at 10 Hz. The laser beam via a reflection mirror is delivered into the scan-
ning probe and redirected by a fixed mirror/prism to the scanning MEMS mirror
(Figure 3.43d). The PVDF ultrasound transducer is attached to the end of the scan-
ning probe (Figure 3.43b). The homemade PVDF ultrasound transducer is a ring-
shaped one-element ultrasound detector. It was fabricated using a 110 µm-thick Ag
ink-printed PVDF film (Measurement Specialties Inc.) as the sensing unit, flexible
PCBs as positive and negative electrode connectors, and backing materials. The
inner and outer diameters of the probe are 9.5 mm and 11.5 mm, respectively. The
central frequency is 2.5 MHz. The photograph of the probe is shown in Figure 3.43c.
As shown in Figures 3.43d and 3.43e, the MEMS mirror has four actuators, which
are folded to obtain a high filling factor (25%). The mirror aperture size is 1 × 1 mm2,
Trigger
Nd: YAG
Trigger Personal computer

Functional
generator
Amplifier and data
aquisition board
Pre-amplifier
(a)
MEMS MEMS
mirror mirror
Transducer
Fixed
mirror
(b) (c)
1 mm
Actuator
(d) (e)
FIGURE 3.43 (a) Schematic of our photoacoustic imaging system. 3D rendering (b) and
photograph (c) of the scanning probe. (d, e) Photograph of the MEMS mirror.
and the device footprint size is 2 × 2 mm2. When applying a voltage generated by two
function generators to the MEMS mirror, each actuator can move up or down due to
the Joule heating effect, resulting in a tilt to the mirror plate. Two-dimensional scan-
ning can be realized by applying different voltages to the four actuators as shown
in Figure 3.43e. In the system, an output trigger signal is used to synchronize the
data acquisition and function generators (AFG3022B, Tektronics). The ultrasound
signal is amplified by a preamplifier with a gain of 17 dB and further amplified by
an amplifier with a controllable gain from 5 to 20. The data acquisition sampling
rate is 50 MHz and a band-pass filter (Panametrics, Waltham, Massachusetts) from
1 to 5 MHz is used to cut off the low-frequency and electromagnetic noises. In the
experiments, ultrasound gel is applied between the probe and the sample to mini-
mize ultrasound attenuation.

We demonstrate the ability of the MEMS-based imaging system through several
phantom/chicken and in vivo experiments. In these experiments, the 532-nm laser
beam illuminated the sample surface over an area of 1 × 1 mm2 at 15 mJ/cm2 which is
below the American National Standards Institute safety limit of 20 mJ/cm2 The four
actuators driven by 0~4 V differential ramp voltages were used for scanning an area of
up to 9 × 9 mm2. The scanning frequencies used were 100 and 2 milliHz along X and
Y directions to collect signals at 50 × 50 points. Due to the 10 Hz repetition rate of the
laser, the scanning time is currently limited to 250 s for the experiments performed.
As shown in Figure 3.44a, the pencil lead was embedded 1.2 mm below the sur-
face of a 50 mm-diameter cylindrical solid phantom, consisting of Intralipid as the
400
1 mm Raw data
300 Hilbert transform
200
100
–100
–200
–300
–400
0 200 400 600 800 1000
(a) (b)
1 mm
(c)
FIGURE 3.44 (a) Photograph of a phantom containing a single object. (b) A typical raw
signal and signal after the Hilbert transform. (c) The recovered 2D image.
1 mm 1 mm
(a) (b)
FIGURE 3.45 Photograph of a phantom containing multiple objects (a), and the recovered
2D image (b).
−1
scatterer and India ink as the absorber. The phantom had a = 0.007 mm (absorp-
tion coefficient) and s = 1.0 mm −1 (reduced scattering coefficient). Agar powder
(2%) was used to solidify the Intralipid/ink suspensions. Figure 3.44b presents a typi-
cal A-line signal collected, along with its amplitude after the Hilbert transform, while
Figure 3.44c shows the 2D image of the object. Using the criteria defined by the
FWHM of multiple A-line signals, the size of the recovered object is estimated to
be 0.72 mm compared to the actual object size of 0.70 mm. The SNR of the image
signals was found to be 32 dB, which decreases with increased depth of the target.
Figures 3.45a and 3.45b give the photograph and recovered image of another
phantom where multiple targets were embedded at different positions/depths. From
the image shown in Figure 3.45b, we see that the targets are all detected.
The results given in Figure 3.46 show the 3D imaging ability of our system. In
this experiment, a pencil lead with a diameter of 0.7 mm was embedded in one piece
1 mm
Target
(b) (d)
(a) (c) (e)
FIGURE 3.46 (a) Photograph of the chicken with an embedded object. (b)–(d) Coronal,
sagittal, and cross-section views of the recovered 3D image. (e) The 3D rendering of the
recovered image.
1 mm
Discontinued part
(a) (b)
FIGURE 3.47 (a) Photograph of the hand with highlighted imaging area. (b) 3D rendering
of the recovered blood vessels.
of chicken (Figure 3.46a) at a depth of 1.5mm from the surface. The recovered 3D

images are presented in Figures 3.46b–3.46d (at coronal, sagittal, and cross-section
views) and in Figure 3.46e (3D rendering of the recovered image). Again, the object
is accurately imaged in terms of its size, shape, and location.
We also demonstrate the in vivo imaging ability of this MEMS-based photoacous-
tic imaging system for visualizing blood vessels in a human hand. The photograph of
the hand and the barely visible blood vessels under the skin are shown in Figure 3.47a,
while the recovered 3D image of the blood vessels is given in Figure 3.47b. We can
see that the blood vessels are clearly imaged with correct shape and size for most
part. We also note that some middle portions of the blood vessels (indicated by arrow
in Figure 3.47b) are notably distorted or missing. This is most likely attributed to
the fact that these portions of the blood vessels are physically located deeper over
other parts of the blood vessels of interest and that the sensitivity of the transducer
responsible for this part of the imaging area is lower due to the limited directivity of
the transducer used.
4 Transducer Array-
Based Photoacoustic
Tomography
2D, 3D, and 4-D
Photoacoustic Imaging
In this chapter, we describe a transducer array-based photoacoustic tomography

(PAT) system and demonstrate its ability for 2D, 3D, and 4-D photoacoustic imaging
of biological tissue.
4.1 ARRAY-BASED PAT SYSTEM AND 2D IMAGING

Figures 4.1a and 4.1b present the photograph of an array-based PAT system. As
shown in Figure 4.1a, a tunable Ti:Sapphire laser (part A) is used to provide 690 to
950 nm near-infrared (NIR) laser pulses with a pulse width of 8 to 25 ns and a repeti-
tion rate of 10 Hz. The photoacoustic signals generated by the laser are received by a
full-ring array with 5 MHz frequency transducers (part B). Preamplifiers/amplifiers
(part C) amplify the signals from the transducers and deliver the amplified signals
to the A/D boards (part D), which are controlled by the computer (part E). The three
key components of the system including the tunable laser, ultrasound transducer
array, and data acquisition system are discussed in detail in Sections 4.1 to 4.3 below.
Figure 4.2 depicts the system block diagram which allows better understand-
ing of the data acquisition flow in this PAT system. The Ti:Sapphire laser optically
pumped by a Q-switched Nd:YAG laser delivers 8 to 25 ns pulses at 10 Hz with
a wavelength tunable from 690 to 950 nm to generate photoacoustic signals. The
192-element transducer array is used to capture the photoacoustic signals. In the
full 192-element configuration, 16 preamplifier boards, each supporting 12 channels,
amplify the signals from the transducer elements with programmable gains of 26 dB.
The multiplexed signal is subsequently amplified with an AD604 amplifier with pro-
grammable gain of 0 to 80 dB, and then digitized with the 8 × 8 channel A/D boards.
The timing for data acquisition is specifically shown in Figure 4.3. We see that the
timer/controller contains a 10 Hz laser Q-switch signal repeated each 100 ms (the
Q-switch pulse width is 100 µs, and the laser pulse width is 10 ns). Three scan enabling
signals, EN_CH1-64, EN_CH65-128, and EN_CH129-192, are synchronized by the
Q-switch signal with a 3.33 Hz frequency (period = 300 ms). EN_CH1-64 enables
109
(E) Computer
(D) A/D
(C) Preamplifier
(F) Tank
(A) Laser
(a) (B) Transducer Array
(b)
FIGURE 4.1 Photograph of the array-based PAT system. (a) Top view. (b) Front view.
the #1~#64 channel preamplifiers and data acquisitions. Each channel has a sample
rate of 50MHz, controlled by signal CLK50M (frequency variable). The post delay
and the sample waveform length are variable and can be assigned by Labview pro-
gram. EN_CH65-128 enables the #65~#128 channels preamplifiers and data acqui-
sitions. EN_CH129-192 enables the #129~#192 channels preamplifiers and data
acquisitions. This provides the opportunity to read all captured pointers into the
Transducer Array-Based Photoacoustic Tomography 111
Ti: Sapphire 16 pre-Amp boards in the box

laser … ..
12 channel /4 output pre-Amp board

Fixed Pre-Amp (26 dB)
5M transducer array (192)

Fixed Pre-Amp (26 dB) MUX (3 to 1)
1-192
…
…
1-64
8 channel AD board
Data Memory A/D Variable gain Amp (0–80 dB) BP Filter
Acquisition
Control & PC 1-64 Memory A/D Variable gain Amp (0–80 dB) BP Filter
Labview
Memory A/D …
Variable gain Amp (0–80 dB) BP Filter
8 AD board in PC
FIGURE 4.2 Block diagram of the real-time PAT system.
computer through the local bus after the acquisition and easy readout of the photo-
acoustic data from all the channels.
4.1.1 Laser Source
A Ti:Sapphire (Symphotics TII, LT-2211A) laser optically pumped with a Q-switched
Nd:YAG laser (Symphotics-TII, LS-2137/2) delivers 8 to 25 ns pulses at 10 Hz with
wavelength tunable from 690 to 950 nm. The beam is expanded with a Galilean
telescope assembly and subsequently diverged with either a plano-concave lens or
100 us
100 ms 100 ms
Q-Switch
Laser 10 ns
CLK50M
EN-CH1-64 Post dealy
EN_CH65-128
Sample waveform length
EN-CH129-192
FIGURE 4.3 Diagram of timing used in the PAT system.

(b) Ultrasound transducers
(a) Full-ring transducer array

(c) Ultrasound signals
FIGURE 4.4 Full-ring ultrasound transducer array.
homogenized by a circular profile engineered diffuser (ED1-S20, ThorLabs, Newton,

NJ). The laser beam is positioned at the center of the transducer array and strikes the
stage orthogonal to the imaging plane for maximum uniformity.
4.1.2 Full-Ring Ultrasound Transducer Array

The transducer array consists of 192 elements arranged in a full circular path (360°)
with a radius of 150 mm (Figure 4.4). The transducer array has a central frequency
of 5 MHz and a bandwidth of greater than 75%. Each element is spaced with a lateral
pitch of 8 mm and has an active area of 3 mm diameter.
4.1.3 Control Electronics and Data Acquisition

The 192-channel photoacoustic data acquisition system is composed of four preampli-
fier/multiplexer boxes and 8 PCIAD850 data acquisition boards (Figure 4.5). Each
preamplifier/multiplexer box includes four boards with 4 × 12 AD8099 chips and 4 × 4
MAX4051 chips. The whole system has 192 AD8099 (26 dB) and 64 MAX4051 chips.
Each PCIAD850 has 8 channels with a sample clock up to 50 MHz and 10 bit A/D con-
verter. The system can sample 64 channels at one time. The #1 PCIAD850 synchro-
nizes with the laser Q-switch sync-signal. The computer sends out the scan enabling
signals by the USB-1024ls.
Each preamplifier (i.e., AD8099) amplifies the signal with a gain of 26 dB
(Figure 4.6a), while MAX4051 is used as a 3 to1 switch having a ±5V DC power
supply (Figures 4.6b). PCIAD850 is an analog-to-digital data converter (two boards)
with simultaneous acquisition ability. Each PCIAD850 supports 8 channels; see
Figure 4.7. At each trigger event (software trigger or external trigger), all the channels
192 transducers
AD8099 ··· AD8099 ··· AD8099

AD8099 AD8099 AD8099
AD8099 AD8099 AD8099
··· ···
MAX4051 MAX4051 MAX4051
··· ···
USB
GAIN AMP GAIN AMP GAIN AMP
1024LS
··· ···
10bitGAIN
A/D AMP 10bitGAIN
A/D AMP 10bitGAIN
A/D AMP
··· ···
GAIN
DSP AMP A/GAIN
10bit PCI AMP
D Bridge 10bit A/GAIN
D AMP
··· ···
··· 10 bit A/D 10 bit
PCIA/D
Bridge 10 bit A/D
··· DSP PCI Bridge

···
(8 PCIAD850 boards) ···
Laser
Computer
Q-switch
FIGURE 4.5 Flow diagram of the 192-channel photoacoustic data acquisition system.
+VS C2 MAX4051
RF 10μF
499Ω NO4 1 16 V+
C3
RG 0.1μF
FB NO6 2 15 NO2
26Ω
– +V COM 3 14 NO1
AD8099 VC VOUT
VIN CC RL NO7 4 13 NO0
+ –V 1kΩ
R1 D
NO5 5 12 NO3
26Ω
DISABLE
C5 INH 6 11 ADDA
0.1μF
V– 7 LOGIC 10 ADDB
C4
10μF GND 8 9 ADDC
DIP/SO/QSOP
–VS
(a) (b)
FIGURE 4.6 (a) 26 dB AD8099 circuit design. (b) Design of the MAX4051 as a 3-to-1 switch.
Block Diagram of PCIAD850/PCIAD1650 Board

Channel 1 Channel 2 Channel 16
Protection Signal in Protection Signal in Protection Signal in
circuit connector circuit connector circuit connector
Pre-amp & Pre-amp & Pre-amp &

10-bit A/D 10-bit A/D 10-bit A/D Control lines to
variable gain variable gain variable gain
converter converter converter next board
amplifier amplifier amplifier
DAC DAC DAC

D/A D/A D/A
memory memory memory
4−32k FIFO 4−32k FIFO 4−32k FIFO
memory memory memory
Bus Bus Bus
transceiver transceiver transceiver
32-bit local data bus
8-bit local data bus
PCI Bus Programmable 50 MHz Power

PCI Bridge DSP chip
configuration logic Clock management
PCI bus
Computer RAM Other

CPU devices
FIGURE 4.7 Block diagram of PCIAD850/1650.
will simultaneously convert the analog signals to digital data with user-selected post-
trigger delay and waveform length. Other programmable parameters include sam-
pling rate, trigger source, trigger rate, gain, DC offset, low- and high-pass filters.
Both boards feature user-selectable 10-bit resolution and up to 50 MB samples per
second. With one board set as the master and remaining set to slave mode, all chan-
nels on the slave boards will start taking data upon receiving a trigger signal from
the master board. The jitter between channels is less than 2 ns. PCIAD850 has low-
pass filters with different cutoff frequency: All frequency, 16 MHz, 7.3 MHz, and
4.8 MHz. It also has high-pass filters with different cutoff frequency: 4.8 MHz,
1.6 MHz, 0.6 MHz, and 0.016 MHz.
Labview programming is used to control the entire data acquisition. The software
is written to allow maximum flexibility for various system configurations. Pointers
created during the data acquisition process are used to quickly access the photo-
acoustic data. Data is then saved for offline image reconstructions. Figure 4.8 shows
the front control panel of the Labview software.
4.1.4 System Evaluation and Experimental Studies

4.1.4.1 System Calibration
The system calibration is primarily concerned with the calibration of the position-
ing and response of the multiple transducers in the system. We used a photoacoustic
(PA) point target centrally located to achieve the goal. In the calibration, the
PA signal was measured by each transducer at all the locations along the circu-
lar path. Figure 4.9 illustrates the measured PA signals from six representative
FIGURE 4.8 Front control panel of the Labview software.
transducers over time-of-flight before and after the calibration. Before calibra-
tion, we see that the signal from the point target arrives each transducer at dif-
ferent time indicating positioning error Figure 4.9). By adjusting the position of
each transducer, we obtained the signals from the point target that arrived at the
transducer at the same time point (red curves in Figure 4.9). Figure 4.10 shows the
reconstructed image of three circular targets before and after the calibration. We
Before calibration
After calibration
0 2 4 6 8 10 12 14 16 18 20
Time/μs
FIGURE 4.9 Measured PA signals from six representative transducers before and after the
calibration.
153.460000 mm 151.490000 mm
8 8
6 2000 6 3000
4 4 2000
2 1000 2
1000
0 0 0
2 2 0
4 1000 4 1000
6 2000 6 2000
8 8
10 3000 10 3000
4000
5 0 5 10 5 0 5 10
Before calibration After calibration
FIGURE 4.10 Reconstructed photoacoustic images of three circular targets before and after
calibration.
see that the calibration provides much-improved image reconstruction in terms of

the target shape and size.
4.1.4.2 Spatial Resolution
To evaluate the system spatial resolution, we constructed a phantom consisting of
two point targets (brass wires; 100 μm in diameter each) suspended in tissue phan-
tom (solidified Intralipid scattering medium). Figure 4.11a shows the reconstructed
image of the two point targets, while Figure 4.11b is the normalized profile of the
reconstructed image along a line y = 10 mm. From the profile, the half-width-at-
half-maximum (AB) and the half-width-at-quarter-maximum (CD) were calculated
to be 0.13 mm. Using the resolution criterion, the spatial resolution of the system was
estimated to be AB+CD–target diameter = 0.13+0.13–0.1 = 0.16 mm.
151.500000 mm, AB = 0.130000 mm,

151.530000 mm
CD = 0.130000 mm, resolution = 0.160000 mm
8.5 4000
4000 3500
9 3000
3000 2500
9.5
2000 2000 CD = 0.13 mm
10 1500 AB = 0.13 mm C D
10.5 1000 1000 A B
500
11 0
0
11.5 1000 500
1000
3.5 4 4.5 5 5.5 6 6.5 3 3.5 4 4.5 5 5.5 6 6.5 7 7.5 8
(a) (b)
FIGURE 4.11 (a) Reconstructed image of two point targets. (b) The normalized profile of
the reconstructed image shown in (a) along y = 10 mm.
152.330000 mm 151.490000 mm
5 25000 8
4 6 3000
3 20000 4 2000
2 15000 2
1000
1 0
10000 2 0
0
1 5000 4 1000
2 6 2000
0
3 8
4 (a) 5000 10 (b) 3000
5 4000
5 0 5 5 0 5 10
FIGURE 4.12 Reconstructed image of a single target (a) and three targets (b).
4.1.4.3 Phantom Evaluation
Phantom experiments were conducted to evaluate the 2D imaging ability of the
array-based system. Figure 4.12a shows the recovered photoacoustic image of a
0.1 mm-diameter target, while Figure 4.12b gives the reconstructed image of three
targets having a diameter of 4 mm, 1.2 mm and 0.6 mm, respectively. All the images
obtained show that these targets with different sizes can be clearly detected using
this array-based imaging system.
Figure 4.13 presents cross-sectional photoacoustic images of a simulated blood
vasculature located at various depths ranging from 0 to 4.6 cm below the phantom
surface. The cross-sections illustrate the capability of the system to provide high-
resolution and high-contrast images at least at a depth of up to 3.5 cm (Figures 4.13a
and 4.13b show the photograph of the samples).
We also conducted phantom experiments to test the high temporal resolution
ability of the system for real-time tomographic imaging of dynamic fluid flow
(3–10 Hz). Figure 4.14 depicts image frames from a 10 sequence of dynamic ink
flow through a 1.6 mm-diameter polyethylene tube using a manual syringe push.
15 On surface 1000 15 1.3 cm 15 2.6 cm 700 A

2000 600
10 800
600
10 1500 10 500
5 400
400 5 1000 5 300
0 200 500 200
0 0 100
0
5 200 0 0
5 5
400 500 100
10 200
600 10 1000 10 300
15 800 15 1500 15
10 5 0 5 10 15 20 10 5 0 5 10 15 20 10 5 0 5 10 15 20
3.5 cm 500 500 4.6 cm 2000 B

4.1 cm
10 1000 10 10 4000
1000
5 1500 5 5 6000
1500
0 2000 0 2000 0 8000
5 2500 5 2500 5
10000
10 3000 10 10
3000 12000
15 3500 15 15
10 5 0 5 10 15 20 10 5 0 5 10 15 20 10 5 0 5 10 15 20
FIGURE 4.13 Cross-sectional images of simulated blood vasculature at various depths.

Laser
Tube
Ink
Experiment diagram
(a)
?? ??
?? ?? ?? ??
?? ??
?? ?? ?? ??
?? ?? ?? ??
?? ?? ?? ??
?? ??
?? ??
?? ??
?? ?? ?? ??
?? ?? ?? ??
?? ?? ?? ??
?? ?? ?? ??
?? ?? ?? ?? ?? ?? ?? ??
?? ??
?? ?? ?? ??
?? ??
?? ?? ?? ??
?? ?? ?? ??
?? ?? ?? ??
?? ??
?? ??
?? ??
?? ?? ??
??
?? ?? ?? ??
?? ?? ?? ??
?? ?? ?? ??
?? ?? ?? ?? ?? ?? ?? ??
?? ??
?? ?? ?? ??
?? ?? ??
?? ?? ??
?? ??
?? ??
?? ?? ??
??
?? ?? ?? ??
?? ?? ??
?? ?? ??
?? ?? ??
??
?? ?? ?? ??
?? ?? ?? ??
?? ?? ?? ?? ?? ?? ?? ??
PA slices during circulation
(b)
0.333000 Second
14 5000
12
10 4000
8 3000
6 2000
4 1000
2
0 0
2 1000
4 2000
6
3000
0 5 10 15
Real-time movie
(c)
FIGURE 4.14 Photoacoustic images of dynamic ink flow through a 1.6-mm-diameter tube.
To avoid partial-view effects and maximize temporal resolution, a sliding window

acquisition was adopted in which element data from the nearest (or equal) time
instant were used for image formation. The images obtained show that we can
clearly track the flow through the tubing with high spatial and temporal resolution.
Such ability of tomographically imaging dynamic fluid flow will provide a power-
ful tool for in vivo real-time monitoring of brain hemodynamics (see Chapter 8).
4.2 3D IMAGING
For 3D imaging, scanning in z direction is needed for a circular or linear array of
transducers, which leads to a nonoptimal elevational resolution in z direction. Two-
dimensional planar array of transducers are easy to implement for 3D PAT imag-
ing; however, due to the limited aperture of 2D planar array transducers, features
with high aspect ratio or with orientations oblique to the transducer surface suffer
from distortion, and the azimuthal resolution is reduced. Compared with a 2D planar
array, a spherical array can offer more complete angular views of the object, provid-
ing both high resolution and accurate feature definition regardless of shape or loca-
tion of the object. Here we describe a sparse spherical array based PAT system that,
for example, can be used for 3D imaging of small animal brains.
4.2.1 System Description
Figure 4.15 depicts the block diagram of the real-time 3D PAT system. A Ti:Sapphire
laser optically pumped with a Q-switched Nd:YAG laser sent 8–12 ns pulses at 10
Hz with a wavelength tunable from 690 to 950 nm. The beam was delivered with
an optical fiber through an opening on the top of the transducer array and produced
16 pre-Amp boards in the box
Ti: Sapphire laser

Optical fiber 12 channel /4 output pre-Amp board
5M transducer array (192)

1–192
Water Fixed Pre-Amp (26 dB) MUX (3 to 1)

tank Fixed Pre-Amp (26 dB)
1–64
Pipe to
hold
the rat 8 channel AD board
Data Memory A/D Variable gain Amp (0–60 dB) BP Filter
Acquisition
Control & PC 1–64 Memory A/D Variable gain Amp (0–60 dB) BP Filter
Labview
Memory A/D Variable gain Amp (0–60 dB) BP Filter
8 AD boards in PC
FIGURE 4.15 Block diagram of the real-time 3D PAT system. The inset is a photograph of
the close-up view of the chamber holding a rat head.
Third 64
Second 64
First 64
(a) (b)
FIGURE 4.16 The spherical transducer array. (a) Photograph of the transducer array. (b) 3D
schematic of the transducer distribution on the interface.
an approximately uniform illumination in a 2 cm-diameter area onto the sample.

The transducer array consisted of 192 transducers placed along a custom-fabricated
white ABS spherical interface containing 610 through holes with counter bores, as
shown in Figure 4.16a. More holes were drilled so that the selection of the trans-
ducer positions on the ball can be flexible. The interface has an outer diameter of
160 mm and an inner diameter of 140 mm, and the diameter of the holes in the ball is
5.7 mm, which fitted well with the transducer (5.5 mm outer diameter). Each trans-
ducer (CUSTOM designed from Blatek, Inc.) has a central frequency of 5 MHz with
a reception bandwidth of greater than 80%. The active area of the transducer is 3 mm
in diameter and the angular acceptance is about 15°. The transducers were glued
onto the interface with epoxy, which can be removed to allow the position change of
the transducers.
There were 16 preamplifier boards separately sealed in four metal boxes, and each
board had 12 input channels, 4 output channels, and 2 digital signal control inputs.
Within each board, 12 dedicated operational amplifier modules (AD8099) individu-
ally amplify the input signals with a fixed gain of 26 dB, and then the amplified sig-
nals were multiplexed into the 4 output channels by 4 multiplexer chips (MAX 4051),
which were controlled by a USB IO digital module (USB-1024LS, Measurement
Computing) through the two digital inputs.
The 64-channel parallel data acquisition system consisted of eight 8-channel PCI
cards (PCIAD850, US Ultratek) in an industrial computer. For each channel, 3000
sampling points were collected at 50 MHz sampling rate in 10 bits, and stored in a
32 k on-board memory before they were transferred to the host machine. Amplifiers
with a programmable gain of 0 to 80 dB along with a 16 MHz low-band-pass fil-
ter were built into the data acquisition system. A Labview program controlled the
data acquisition, and the acquired data were stored on hard disk for further image
processing. Images were reconstructed with the delay-and-sum algorithm described
in Chapter 1. For 3D image display, the reconstructed results were normalized to
0~255 after setting the negative values to be zero. Then 3D images were rendered
with Amira (from Visage Imaging, Inc.) with different thresholds as indicated in the
colorbar shown in each image.
The system allows the selection of transducer positions on the spherical interface.
The total 610 holes formed 11 evenly spaced layers along the vertical direction of
the ball, and the 192 transducer positions were indicated with three different colors
in Figure 4.16b. Three different colors were used to indicate the 3 to 1 multiplexing
from the 192 transducers to the 64-channel data acquisition. This system can also be
used as a real-time 2D system operating at 10 f/s using the 64 transducers arranged
in the vertical center layer (blue). For in vivo experiments, the rat head was elevated
to the center of the spherical interface through a chamber fixed at the tank bottom,
whose top was about 15 mm beneath the interface center, and a transparent plastic
wrap was used to cover the chamber top. For phantom experiments, a homemade
silicone holder was used to hold the phantom.
4.2.2 Phantom Experiments
Three different types of phantoms were used: One containing a point object for
system calibration, one phantom containing three hairs tilted along different ori-
entations for static imaging, and one phantom with ink flowing through a thin tube
embedded in a phantom for real-time 3D imaging.
The point object used for calibration was a small spherical graphite particle
(0.1 mm in diameter) located at the center of the spherical array and ensured an
isotropic acoustic emission profile for all directions. We measured and compensated
the delay of time for all the 192 channels in the radical direction, and reconstructed
this point object after calibration. The hairs-containing phantom was used to demon-
strate the high imaging quality of our system. Finally, we imaged an embedded tube
filled with flowing ink to show the real-time imaging ability of our system. The tube
had a 0.3 mm inner diameter and was horizontally placed in a phantom. No averag-
ing of signals was performed for the phantom experiments except for the point object
experiment where 10 times averaging was applied.
We calibrated the system by recording the emission profiles from the spherical
graphite particle for all the 192 channels, and then measured and compensated the
time delay of each channel. We then evaluated the system resolution by reconstruct-
ing the image of the point object.
Figures 4.17a and 4.17c present the reconstructed x–y and z-x cross-section
images of the point object located at the array center. The quality of these images
is determined by both the distribution and the characteristics of the transduc-
ers. The profiles of the two reconstructed images were also extracted in x and
z directions, as shown in Figures 4.17b and 4.17d, respectively. The FWHM of
the profiles was measured to be 0.19 mm (x direction) for Figure 4.17b, and
0.27 mm (z direction) for Figure 4.17d, compared to the theoretical value of
0.16 mm for the 5 MHz central frequency transducer with an estimated cut-off
frequency of 7 MHz. It is noted that the profile in Figure 4.17d is noisier than
that in Figure 4.17b. This along with a larger FWHM from Figure 4.17d was due
to the asymmetric distribution of the transducers. For targets located away from
the center of the array, the radial resolution will stay nearly the same as that for
a centrally located target, while the lateral resolution will be linearly reduced
with increased distance away from the array center. In our system, the lateral
0.5 200
200
100 100
0
0 0
0.5
0.5 0 0.5 0.5 0 0.5
(a) (b)
0.5 200 200
0 100 100
0.5 0 0
0.5 0 0.5 0.5 0 0.5
(c) (d)
FIGURE 4.17 (a) and (c) x–y and z-x cross-section images through the center plan of the
point object. (b) and (d) The profile extracted in x and z directions from (a) and (c), respec-
tively. Units are in mm.
resolution will be reduced by 0.1mm when the target is located 5mm off the
array center.
Figure 4.18a is the photograph of the phantom containing three hairs tilted along
different orientations, and Figures 4.18b and 4.18c show the reconstructed 3D images
from two different views. The reconstructed volume is 10 × 10 × 10 mm with a 0.1
mm voxel size. The spatial distribution and tails for all the three hairs were clearly
revealed. This result indicates that our system is capable of 3D imaging small objects
of different spatial distribution and orientation in high quality.
(a) (b) (c)

150
30
FIGURE 4.18 (a) photograph of the phantom containing three tiled hairs; (b)–(c) recon-
structed 3D images of the three hairs in two different views. Scale bar represents 5 mm.
(a) (b) 11.7 S (c) 12.0 S (d) 12.3 S (e) 12.6 S
220
(f) 12.9 S (g) 13.2 S (h) 13.5 S (i) 13.8 S (j) 14.1 S
70
FIGURE 4.19 Reconstructed 3D images of ink flowing through a 0.3 mm tube embed-
ded in a background phantom. (a) Photograph of the phantom containing the tube. (b)–(j)
Reconstructed 3D images at different time points. The time interval is 0.3 s. Scale bar rep-
resents 5 mm.
The reconstructed 3D images of ink flowing through a thin tube are shown
in Figure 4.19. The image domain is 15 × 5 × 10 mm with a 0.1 mm voxel size.
Figure 4.19a is the photograph of the phantom containing a tube filled with ink, and
Figures 4.19b–4.19j are the reconstructed 3D images at different time points. The
time interval between two consecutive images was 0.3 s. These 3D images clearly
tracked the flow through the tube over the course of 2.4 s with high spatial and tem-
poral resolution, and the flowing speed of the ink was measured to be 6 ± 0.9 mm/s
from the reconstructed results with a time interval of 0.3 s.
4.2.3 In Vivo Experiments

Epilepsy is a serious brain disorder involving intensive hemodynamic changes, which
provides high endogenous contrast for PAT imaging due to the strong absorption of
blood at visible and NIR wavelengths. Compared with the existing neuroimaging
methods (such as MRI, CT, PET, and SPECT), PAT provides not only high ultrasound
resolution and high optical contrast, but also unprecedented advantage of high tem-
poral resolution over these methods, which is critical for capturing seizure dynamics.
Two small rats (~40 g) were imaged with intact skull and skin, but hairs on the
head were removed. The rats were anesthetized and mounted on the homemade
plastic chamber/holder. Focal seizure was induced by microinjection of 10 μl of
1.9 mM bicuculline methiodide (BMI) into the neocortex of one rat, while saline
solution was injected into the brain of another rat as control. In each experiment,
the rat was elevated to the transducer array center and kept alive under the water
tank through the whole experiment. The incident energy of the 730 nm light was
maintained at 8 mJ/cm2 below the safety standard. Seizure process was recorded for
50 min, and the measurement from the control rat was recorded for 3 s. All animal
procedures were performed in accordance with the approved University of Florida
IACUC protocols.
Figure 4.20b presents the 3D images for the rat with BMI injection during the sei-
zure onset at six time points, compared with that for the control rat in Figure 4.20d.
(b) 30.3 s 30.6 s 30.9 s

200
(a)
BMI Injection
31.2 s 31.5 s 31.8 s
20
(d) 0.6 s 0.9 s 1.2 s

200
(c) 3
1.5 s 1.8 s 2.1 s
20
FIGURE 4.20 (a) and (c) Photograph of the rat with BMI injection and the control rat after
scalp removed. (b) and (d) 3D PAT images at six time points for (a) and (c), respectively. The
three main blood vessels are indicated by the white arrows, and the seizure focus is indicated
by the circle in (b). The time interval between two successive images is 0.3 s. Scale bar rep-
resents 10 mm.
The corresponding photographs of the two rats with scalp removed right after the
experiments are shown in Figures 4.20a and 4.20c, respectively. The reconstructed
domain for the images shown in both Figures 4.20b and 4.20d is 20 × 20 × 4.5 mm,
with a 0.1mm pixel size. The three main blood vessels on the rat brain are clearly
revealed for both cases, as indicated by the white arrows, and for the rat with BMI
injection a seizure focus can be clearly seen right at the BMI injection position (the
white circle in Figure 4.20b), where strong absorption is observed during seizure
onset. The rapid changes of the absorption both in the main blood vessel and in the
seizure focus were observed from Figure 4.20b, while no such changes were noted
for the control rat (Figure 4.20d). The seizure focus had a diameter of ~3 mm. This
experiment demonstrates that our system can be used to investigate the hemodynamic
changes in small animal brain both spatially and temporally during seizure onset,
although the complex microvasculature cannot be resolved due to the limited number
of transducers used here. The microvasculature can be revealed when the 192-ele-
ment full-ring array is used for 2D tomographic imaging of rat brains (see Chapter 8).
4.3 4-D IMAGING
Real-time 3D or 4-D imaging is necessary in many cases. Real-time 3D imaging
of epilepsy shown in Section 4.2 is one example. Four-dimensional PAT can also
play a significant role in the localized drug delivery using metal needles. Accurate
Time
t θ1 t θ2
Space
θ1
z
θ2
θ3
θ y
α
θ4
x
θ5
FIGURE 4.21 Schematic representation of time-resolved 4-D photoacoustic tomography.

A series of 2D images at various projection angles and time steps are taken to construct the
tomograms.
and updated knowledge of the position and orientation of the needle can increase
the needle guidance and placement within the tissue. Four-dimensional PAT can
be used to guide photothermal therapy for monitoring temperature distribution
during the therapy. In all these cases, real-time 3D photoacoustic imaging is
needed to capture the physiology and pathology dynamically in three dimensions.
In 4-D PAT, the time resolution as the fourth dimension is integrated with the
3D spatial resolution obtained from a series of 2D projections of an object. Here
we demonstrate the methodology in needle based drug delivery guidance, with a
temporal resolution of 330ms, thus enabling the visualization of dynamic biodis-
tribution and pharmacokinetics of indocynine green (ICG). We also demonstrate
4-D PAT for imaging dynamic changes of temperature during thermal therapy of
tumor.
The 4-D PAT imaging system is described in Section 4.2.1. The concept 4-D
PAT is illustrated in Figure 4.21, which depicts the construction of tomograms
from 2D projections at different angles and times. Because of the various dimen-
sions involved, we note that at a given time each 2D projection represents a 3D
frame (including time), whereas a 3D tomogram when constructed from all the 2D
2.331 s 5.328 s 9.330 s 9.657 s 9.990 s
FIGURE 4.22 PAT-guided curved needle insertion and drug delivery. From left to right, the
needle was imaged over time from 2.331 to 9.990s.
projections represents a 4-D frame. For the reconstructed images shown in this sec-
tion, z scanning from −15 mm to +15 mm with an increment of 100 µm was used, and
the time scale t ranged from 330 µs to tens of minutes.
4.3.1 Drug Delivery Monitoring

A series of 3D snapshots of PAT-guided curved needle insertion is shown in
Figure 4.22. As the 27 G needle was inserted into a tissue-mimicking phantom, it
was photoacoustically tracked in real time. From left to right, the needle was imaged
at different time points from 2.331 to 9.990 s. The needle was remarkably imaged
with excellent contrast (11.5 ± 5, standard deviation) with respect to the background
medium. Both the curved needle progression and the drug (ICG) delivery (at t =
9.657 and 9.990 s) could be detected clearly. The ICG pharmacokinetics in rat brain
could also be monitored in real time (Figure 4.23). Here we present 4-D tomograms
of the rat brain at two representative time points (t = 0.333 s, Figure 4.23a; t = 12.978 s,
Figure 44.23b). Immediately after injection, ICG accumulation in the cortex was
observed photoacoustically. The PAT image acquired at 12.978 s postinjection shows
that the ICG accumulation enhanced the PA signal in the marked area B by 8.5 + 2.6
(standard deviation). After transient accumulation of ICG, it began an exponential
decay of the PA signal as ICG was washing out. The 2D PAT image extracted from
the 3D tomogram of the rat cortical vasculature (Figure 4.23c) matched well with the
anatomical photograph (Figure 4.23d) obtained after the PAT imaging. The brain
structures including the middle cerebral artery, right hemispheres, left hemispheres,
left olfactory bulbs, and right olfactory bulbs are clearly shown in the PAT image.
Numbers 1–3 marked in the image indicate that the micro blood vessels with a diam-
eter of less than 100 μm are also seen, which again correspond well with the rat brain
photograph.
4.3.2 Tumor Therapy Monitoring

Four-dimensional photoacoustic imaging was also used to monitor tissue tempera-
ture variations during tumor photothermal therapy on mouse tumor. Figure 4.24a
(a) t = 0.333 s (b) t = 12.978 s
A B
y
z
13 243 13 243
(c) 350 (d)

300 1
9
1
250
2
2 200
3
3 150
100
8
0.2 mm 50
FIGURE 4.23 4-D visualization of ICG-based pharmacokinetics of drugs. Representative

time frames of 3D volume images of rat brain vascular structures taken at t = 0.333 s (a) and
(b) t = 12.978 s (b). The image acquired at 12.978 s postinjection shows that the ICG accu-
mulation enhanced the PA signal in the marked area B by 8.5 + 2.6 (standard deviation). (c)
The 2D noninvasive PAT image extracted from the 3D tomogram of the rat cortical vascu-
lature. (d) Open-skull photograph of the rat brain surface acquired after the PAT experiment.
Numbers 1–3 indicate the corresponding blood vessels in the PAT image and rat brain pho-
tograph. Male Sprague–Dawley rats (Harlan Labs, Indianapolis, IN) weighing 50–60 g were
used in this study. ICG in phosphate buffered saline (pH 57.4) was injected intravenously at a
dosage of 0.25 ml/100 g body weight, which produced an estimated ICG blood concentration
of 131,025 M. After injecting ICG, a series of PA images was obtained to dynamically moni-
tor ICG uptake in the rat brain.
1–6 shows representative 2D slices and Figure 4.24a7 is 3D reconstructed photo-

acoustic image of mouse mammary tumor at 810nm light illumination. Using the
optical absorption difference between tumor and surrounding normal tissue, the
tumor margins can be easily identified (Figure 4.24a1–6) and corresponds well with
tumor histology (Figure 4.24a8).
In the experiments sliver nanoparticles were used to effectively enhance pho-
tothermal therapy because of their high optical absorption. Thirty minutes after
sliver nanoparticle injection, photothermal therapy was performed for 5 minutes
using a CW laser at 755 nm. During therapy, photoacoustic signals at pulsed laser
0 min 1.2 min
1 2
0.4 min 1.6 min 90 s 120 s
∆T (°C)
4
0.8 min
150 s 180 s
(b)
5 6
6
Maximum ∆T (°C)
4
7 8
(a)
2
0
0 50 100 150 200 250
Time (s)
(c)
FIGURE 4.24 Tumor margin detection and thermal therapy monitoring. (a) Three-
dimensional stack of consecutive photoacoustic image sections from a breast tumor on
mouse; the 0.2 mm slice corresponds to sections 1–5. Maximum intensity projection of the
entire z stack is shown in 6; 7 is the 3D reconstruction of the entire z-stack. Yellow dashed
circle represents the tumor margin in different slices as in 1–6 which correlated well with
tumor histology in 8. (b) 3D photoacoustic-based thermal images at different time points (90,
120, 150 and 180s) after photothermal therapy. (c) Photoacoustic measurement of temperature
rise during photothermal therapy; a maximum temperature rise of ~6.6ºC within the tumor
is seen. Error bars (±s.d.) were calculated from five consecutive measurements. In this study,
1-month-old BALB/C mice were injected in the right flank with EMTCL2 human cancer cells
(about 53,106 cells) and tumors were allowed to grow for 15 days to a size of 10 mm in diam-
eter. Sliver nanoplates were used for thermal therapy due to their enhanced thermal stability
and enhanced photoacoustic signal response. By precisely controlling the plate diameter and
thickness, the nanoplate’s optical resonance can be tuned to peak at specific wavelengths at
785 nm. The injected volume of nanoparticle solution was 200 mL, and the injected concen-
tration was 0.08 mg/ml. The nanoparticles were intradermally injected at the tumor site. A
CW laser (785 nm, 80 mW, Model DL7140-201S, Sanyo, Japan) was used for photothermal
therapy. Prior to taking experimental measurements, tissue temperatures were measured by a
thermistor to calibrate the relationship between temperature and photoacoustic signal ampli-
tude. During experiments, tissue samples were immersed in a water tank maintained at 37ºC,
and thermistor was inserted into the center of tissue sample to measure the temperature.
wavelength of 810 nm were collected. The CW laser beam spot size on the skin of
the mouse was 0.8 mm in diameter. Based on temperature calculation models (Shah
et al. 2008), a temperature distribution was estimated. A maximum temperature
rise of 6.6ºC during therapy was observed by photoacoustic-based thermal imag-
ing (Figure 4.24c). During the first 1 minute 40 seconds, mouse tumor temperature
gradually increased, and then the temperature was relatively constant until the CW
laser was turned off.
5 Photoacoustic
Microscopy and
Photoacoustic
Computed Microscopy
Photoacoustic microscopy (PAM) is just a variation of photoacoustic tomogra-

phy (PAT). In general, PAT uses relatively low frequency (up to 5–7 MHz) unfo-
cused transducer(s) to realize deep tissue imaging with the sacrifice of reduced
spatial resolution, whereas PAM is based on high frequency (>10 MHz) focused
transducer(s) to achieve high spatial resolution with the sacrifice of reduced tissue
penetration. Depending upon the spatial resolution and applications, PAM is clas-
sified into several categories including optical-resolution PAM (ORPAM), acous-
tic-resolution PAM (ARPAM), C-scan PAM, photoacoustic computed microscopy
(PACM), PAM based on acoustic lens with variable focal length, and confocal
PAM using a single multifunctional lens. Here we describe the principles of these
PAM approaches.
5.1 OPTICAL-RESOLUTION PHOTOACOUSTIC MICROSCOPY

ORPAM uses optical confinement for localization purpose and is similar to many
conventional optical microscopy techniques where the lateral resolution is deter-
mined by the dimension of a focused light spot. Due to high optical scattering in most
tissues, the imaging depth is limited to be less than 1 mm for ORPAM. However, it
has the highest lateral resolution of 5 µm or better compared with that of ARPAM
or PAT. ORPAM has shown to be a powerful tool especially for label-free in vivo
imaging of microvasculature, and thus provides a valuable complement to the exist-
ing microscopy techniques.
As shown in Figure 5.1, an ORPAM system employs optically focused laser
beam from a pulsed Nd:YAG laser to achieve microscale lateral resolution. Laser
pulses with duration of 6 ns and repetition rate of 10 Hz are spatially filtered by a
50 µm-diameter pinhole. Then the laser beam is collimated by a convex lens and
focused by an objective lens to obtain a focal spot of 5 µm in diameter laterally.
The laser energy at the tissue surface is 25 mJ/cm2, which is less than the skin dam-
age threshold. A 50 MHz focused transducer with 3 mm activate area and 6 mm
focal length is confocally positioned with the focused light beam. As described in
131
Transducer
2D linear stage
Water tank
Optical window
Objective lens
Light beam
FIGURE 5.1 Schematic of ORPAM system.
Chapter 1, the volumetric photoacoustic images can be generated by raster scanning

the laser beam/transducer along the x–y plane coupled with extraction of depth-
resolved photoacoustic signals.
To demonstrate the imaging ability of this ORPAM system, we employed it to in
vivo visualize the microvasculature in an ear of a nude mouse (body weight: 25g)
with excitation source at 532 nm. The animal procedures were conducted in confor-
mity with the laboratory animal protocol approved by Animal Studies Committee
of University of Florida. Before the experiments, the mice were anesthetized with a
mixture of ketamine (85 mg/kg) and xylazine. An imaging area covering 1 × 1 mm2
was scanned with a step size of 6µm without any signal averaging.
In Figure 5.2a, a typical microvasculature of mouse ear imaged by ORPAM is
shown where we can easily identify single capillaries with a size of less than 5 µm.
As indicated by arrow, red blood cell (RBC) can be seen among these capillaries as
well. In Figure 5.2b, the top panel shows the ORPAM image of two blood vessels in
comparison with that by a transmission optical microscope at a 4 × magnification
(bottom panel). These two blood vessels are clearly identified by ORPAM, whereas
they cannot be seen by the transmission optical microscope.
5.2 ACOUSTIC-RESOLUTION PHOTOACOUSTIC MICROSCOPY

ARPAM comprises a focused transducer and a weakly focused light beam where
excitation light is delivered through a conical lens to realize annular or dark-field
illumination with a dark center, or without the use of such a conical lens to achieve
Photoacoustic Microscopy and Photoacoustic Computed Microscopy 133
50 µm
250 µm RBC
(a)
(b)
FIGURE 5.2 (a) Maximum amplitude projection (MAP) photoacoustic images of mouse
ear. (b) Two blood vessels imaged by ORPAM (top panel) in comparison with that by conven-
tional transmission microscope image (bottom panel). Note that the blood vessels cannot be
identified by transmission optical microscope; the two red lines are used to help the visualiza-
tion of the position of the blood vessels.
direct or bright-field illumination without a dark center. Although the lateral resolu-
tion of ARPAM is determined by the bandwidth of the transducer (typically from
50 μm to 1 mm) and lower than ORPAM, it can detect target more deeply located
than that by ORPAM. Because of this, ARPAM has found more applications than
ORPAM (see Chapter 8 for details).
Figure 5.3a shows the schematic of a typical ARPAM system. In this system,
a short-pulsed laser beam of 6 ns duration at 10 Hz repetition rate from a 532 nm
Nd:YAG laser is divided into two beams by a beam splitter and coupled into two
optic fiber bundles. Wideband photoacoustic waves induced as a result of ther-
moelastic expansion of target is collected by a focused high-frequency ultrasound
transducer (50 MHz) with 3 mm aperture and 6 mm focal length. The optical illu-
mination through the optic fiber bundles is adjusted to optimize the signal-to-noise
ratio (SNR) of the system as shown in Figure 5.3b. The laser beams are delivered to
the target through a water tank with a window sealed with an optically and ultra-
sonically transparent membrane. The acoustic transducer and optic fiber bundles are
raster scanned by a 2D moving stage with a scanning step size of 60 μm. The image
formation method is the same as that for ORPAM (see Chapter 1).
For the purpose of demonstration, the vasculature of rat ear is imaged in vivo
using ARPAM. Figure 5.4 presents the maximum amplitude projection (MAP) pho-
toacoustic image projected along the vertical direction (z-axis) to the orthogonal
plane. We see that the vasculature of the rat ear is clearly identified.
Beam splitter
(a) Lens
Nd: YAG laser
PC Lens
Fiber bundle II
Amplifier
Fiber bundle I Transducer
2D
scanner
(b)
Illumination
Tank pattern
Membrane
Active area of Holder
transducer (3 mm)
FIGURE 5.3 (a) Schematic of an ARPAM system. (b) Illustration of optical illumination
pattern and active acoustic area.
1 mm
FIGURE 5.4 In vivo MAP photoacoustic image of vasculature in a rat ear.

5.3 C-SCAN PHOTOACOUSTIC MICROSCOPY

We have shown the high-resolution imaging ability of ORPAM and ARPAM for
visualizing high-absorption structures such as blood vessels. For mesoscopic-scale
animals such as Drosophila, zebrafish, and Xenopus laevis, the structures inside
these animals do not absorb light strongly. In addition, it is known that the use of
low-power illumination is a required condition for imaging these animals because
high-power illumination will either interrupt the biological development of interest
or damage the animals. Without signal averaging, the SNR for low-absorption tar-
gets or under the condition of low-power illumination will not be able to produce a
usable image for conventional PAM. Another limitation associated with the conven-
tional PAM is its limited depth of focus for the focused high-frequency transducer
used, which generates deteriorated images for out-of-focus planes. Here we describe
a new PAM approach based on axial scanning or C-scan instead of signal averaging
to improve the out-of-focus resolution for low-absorbing objects under the condition
of low-power illumination.
5.3.1 Materials and Methods

5.3.1.1 Experimental Setup
In our C-scan PAM (CPAM) system shown in Figure 5.5a, a 532-nm pulsed laser
is used to generate a short-pulsed laser beam with 6-ns duration at 10 Hz repetition
(a) Beam splitter

Lens
Pulsed laser
PC Lens
Fiber bundle II
Amplifier
Fiber bundle I Transducer
(b) Focused
transducer
3D
scanner
Start
2 position
Focal point X Y
4
3
End position Holder Z
FIGURE 5.5 (a) Schematic of the PAM system. (b) C-scan PAM concept. The start and end
positions are typically 0.5 mm away from the focal point.
rate. Then the laser beam is split into two subbeams and each of the subbeams is
coupled into an optic fiber bundle. The two optic fiber bundles are fixed in opposite
positions to produce a homogenous volume illumination through the object. Induced
photoacoustic waves are received by a focused high-frequency ultrasound transducer
(50 MHz, 3 mm aperture, and 6 mm focal length). The imaging probe consisting of
the transducer and optic fiber bundles is mounted on a 3D moving stage. This system
can provide a lateral resolution of 61µm on the focal point (Figure 5.5b) and an axial
resolution of 15 µm.
In the experiments, a Drosophila pupa was fixed on a holder with its long side
parallel to the y-axis (Figure 5.5a). The scanning step was set to be 25.4µm along the
x- and y-axis and 50 µm along the z-axis. During the in vivo experiments, low laser
power (0.5 mW) was used to ensure uninterrupted biology development of the pupa.
We noticed that all the pupae imaged developed into fruit flies with this laser power.
5.3.1.2 Image Reconstruction Method

In conventional PAM, photoacoustic signal is collected at each location of trans-
ducer, resulting in (1-D) depth-resolved image. A 3D image is produced by scan-
ning the ultrasound transducer in X–Y plane. For low-optical-absorbing objects, the
photoacoustic signals need to be averaged at each transducer location to increase the
SNR. In our CPAM, for each 1-D depth-resolved image the focal point of the trans-
ducer is scanned along the z-axis (i.e., axial scanning from the start position to the
end position in Figure 5.5b) and the signals collected at all the points along the z-axis
are used to reconstruct a 1-D image as follows:
j =1
∑ p( j, t − t )
n
0
P(t ) = (5.1)
n
where p(j, t) is the signal received at the j the focal point during the axial scanning,
t0 is the calculated time of acoustic wave propagation for one scanning step along the
z-axis (t0 = 0.03 µs in this study), and n is the total scanning steps used for the axial
scanning.
5.3.1.3 Phantom Preparation
The image reconstruction method described above was validated through phantom
experiments. In these experiments, a human hair was embedded in an agar powder
(2%) solidified tissue mimicking phantom, consisting of Intralipid as the scatterer
and India ink as the absorber. The absorption coefficient of the phantom was 0.007 mm–1
and the reduced scattering coefficient was 1.0 mm–1. Due to the high absorption of
hair, we used a very low power illumination of 0.1mW for these experiments.
5.3.1.4 Animal Preparation and Histological Sectioning

Drosophila melanogaster of the genotype elav-Gal4; UAS-DsRed were grown in
standard media in 25ºC. Within 24 hours following pupation, pupae were removed
from the vial. A small hole was made at the posterior as well as the anterior end of
the pupal case with a fine tungsten needle under a dissection scope. The punctuated
pupae were quickly immersed in 1 ml of 4% paraformaldehyde in PBS (phosphate
buffered saline) in a 1.5 ml eppendorf tube. The tube was let rotate in room tempera-
ture for 24 hours. Certain histological sections were performed by frozen sectioning
machine and imaged by fluorescence optical microscopy system.

The image probe was scanned for 25 steps with a step size of 50 µm along z-axis
from starting position to ending position (Figure 5.5b) and the raw data collected
from all the axial-scanning points were used to produce an A-line image. Each 2D
image contained 50 A-line images. Figure 5.6a (T1, upper left) shows the recon-
structed image using the axial-scanning method, while Figure 5.6a (T2, T3, and T4)
show the images recovered by averaging the signals 25 times without the use of the
axial scanning when the hair was in positions 2, 3, and 4, respectively (Figure 5.5b).
T1 1 T2 2 T1
1' 2'
100 µm 5 6 7
T3 3 T4 4 T2
3' 4'
5' 6' 7'
(a) (b)
3
2.5 Line 1 Line 1'
Line 2 2.5 Line 2'
Photoacoustic Signal
2 Line 3 Line 3'

Line 4 2 Line 4'
1.5 1.5
1 1
0.5 0.5
0 0
–0.5
100 200 300 400 500 0 150 300 450
x-axis (µm) z-axis (µm)
(c) (d)
FIGURE 5.6 (a) The image of hair by C-scan PAM (T1), and the image of hair located at
different positions by conventional PAM (T2, T3, T4). (b) In vivo pupa images by C-scan
PAM (T2) and by conventional PAM (T1). (c) Lateral profiles for the images shown in Figure
5.6(a). (d) Axial profiles for the images shown in Figure 5.6(a).
TABLE 5.1
Comparison of the Target Size Measured by CPAM and Actual Value
Recovered Recovered
Distance from the Actual Size Lateral Size Axial Size
Line # Focal Point (um) (um) (um) (um) Error (lateral)
1 NA 98 105 99 6.1%
2 450 98 170 99 73.5%
3 300 98 135 99 37.6%
4 0 98 99 99 1.0%
Transverse and axial profiles along the dashed lines 1, 1’, 2, 2’, 3, 3’, 4, and 4’ in
Figure 5.6a are given in Figures 5.6c and 5.6d after Hilbert transform of the raw data.
Quantitative information was extracted from the profiles shown in Figures 5.6c and
5.6d and summarized in Table 5.1.
From Table 5.1, we see that T1 image provides accurate quantitatively recon-
structed transverse hair diameter with an error of 6.1% compared to the true value.
T2 and T3 images show a large error of 73.5% and 37.6%. T4 image gives the most
accurate transverse size of target with an error of less than 1% because the hair was
placed right at the focal point of transducer. For axially imaged size, accurate results
with an error of less than 1% were obtained for all four cases since the axial reso-
lution (15 µm) was determined by the frequency and bandwidth of the transducer
regardless of the target positions.
5.3.2.2 In Vivo Experiments

For embryos and larvae, the morphogenesis of the central nervous system (CNS) can
be monitored with conventional fluorescence microscope (FM). However, the signal
is completely blocked following the formation of the pupal case using conventional
FM. The fragileness of the pupal tissue coupled with the hardness of the pupal case
make it extremely difficult to dissect the animal without harming the normal devel-
opmental process. With the CPAM technology, it is now feasible to in vivo observe
the development and morphogenesis of the CNS in the pupal stage.
The in vivo PAM experiments were conducted following the same procedure as
the phantom experiments except for the use of increased laser power (0.5mW). The
pupa was fixed in a small tube filled with transparent agar powder solidified mate-
rial as shown in Figure 5.7a. The scanning took 2 minutes for each 2D PAM image.
Figure 5.6b presents the PAM images of pupa by C-scan PAM (T2) and by conven-
tional PAM (T1). The DsRed-expressing cells indicated by arrow 5 in T1 are more
accurate than the same part indicated by 5’ in T2 because two cell clusters are dif-
ferentiated in T1 while they are not distinguishable in T2. In this case these DsRed-
expressing cells were close to the focal point so that we could obtain the highest
4
2 1 3
1
2
5
b 100 µm
c d
e
2 4
1
2
3
1 5
(a) (b) (c)
3 4
1
1
5
2
2
1
3
1 5
(d) (e )
FIGURE 5.7 In vivo CPAM images (top row, b–e), and fluorescence microscopic images of
histologic sections (bottom row, b–e).
resolution. For other DsRed-expressing cells indicated by 6, 6’, 7, and 7’, we also see
that the image quality with C-scan is notably higher.
Figure 5.7a shows the position/section corresponding to the CPAM/FM slices
presented in Figures 5.7b to 5.7e, where the top row shows the in vivo photoacous-
tic images and the bottom row gives fluorescence microscopic images of the cor-
responding histological sections. Major features are clearly identifiable in CPAM
images and agree well with the corresponding histological sections.
To validate if longitudinal observations could be possible, the morphogenetic
movements that occur in the pupae were monitored in vivo by our CPAM as shown
in Figure 5.8. We monitored the same position of one pupa during the first 3 hours
after pupariation. The time-lapse CPAM imaging sequence shows the movements of
DsRed-expressing cells inside the pupa with high resolution.
0:00 0:30
100 µm
(a) (b)
1:30 3:00
(c) (d)
FIGURE 5.8 In vivo CPAM images over time.
5.4 PHOTOACOUSTIC COMPUTED MICROSCOPY

PAM has been shown so far to be particularly useful for imaging microvasculature in
high resolution. PAM, however, is essentially qualitative in nature since it measures
the absorbed optical energy density, the product of the absorption coefficient, and
the local optical fluence rather than the absorption coefficient itself. This prevents
PAM from quantitatively measuring important functional parameters including oxy-
hemoglobin (HbO2), deoxyhemoglobin (HbR), and oxygen saturation (sO2). Absolute
quantification of these functional parameters will allow for correct determination of
the physiological status of tissue and accurate diagnosis of pathological conditions
such as tumors and neurodisorders.
Here we describe a new photoacoustic microscopy method, which we term photo-
acoustic computed microscopy (PACM), to address the critical need for a truly quan-
titative photoacoustic imaging in microscale. Central to PACM is a model-based
inverse reconstruction algorithm, which uses the photoacoustic signals measured
by a focused transducer to obtain quantitative images of HbO2, HbR, and sO2. In
addition, assisted with an oxygen-transport model, PACM permits determination
of two other important functional parameters, blood flow (BF) and rate of oxygen
metabolism (MRO2). These two parameters currently are available from macro-
scopic imaging techniques such as positron emission tomography (PET) and single
photon emission computed tomography (SPECT). PET and SPECT, however, use
radiotracers and cannot image hemodynamics associated with microvasculature due
to their poor spatial resolution.
Nd: YAG laser Coupler

(a)
Ti:Sapphire laser
Mirror
Fiber bundle I
PC
Coupler
Amplifier
Transducer
Function
generator Fiber bundle II
3D scanner
(b)
Illumination
Tank pattern
Water tank
Membrane
Active area of
transducer (3 mm)
FIGURE 5.9 (a) Schematic of the PAM system. (b) Top view illustration of the optical illu-
mination area.
5.4.1 Methods
5.4.1.1 PAM Imaging System
The PAM imaging system is schematically shown in Figure 5.9a. Short-pulsed laser
beam at 10 Hz repetition rate from a Nd:YAG laser and/or a Ti:Sapphire laser can
be used for imaging. The laser beam from each laser is delivered to the sample via
an optic fiber bundle. Both of the optic fiber bundles are positioned to produce an
optimal illumination at the sample surface as shown in Figure 5.9b. Depending on
the application, either one of the lasers or both of them are used for multispectral
imaging (e.g., 1064 nm from Nd:YAG and 730nm from Ti:Sapphire) where a 50 µs
time delay between the two lasers provided by the function generator is used to avoid
signal overlap. The imaging probe consisting of a focused transducer and the optic
fiber bundles is mounted on a 2D moving stage. The sample is positioned under
an open imaging window at the bottom of the water tank sealed with an ultrasoni-
cally and optically transparent membrane. Ultrasound gel is used between the mem-
brane and the animal tissue/phantom for acoustic transmission. At each imaging
probe position, the laser-induced photoacoustic signal is received by the transducer,

amplified by the amplifier and digitalized by a 8-bit data acquisition board (NI5152,
National Instrument) at a sample rate of 250 MS/s. One- or two-dimensional raster
scanning of the imaging probe along the horizontal plane coupled with the depth-
resolved ultrasonic detection forms a 2D or 3D photoacoustic image. It takes ~6s for
a typical B-scan 2D imaging. For the phantom experiments, a focused transducer
with 3.5 MHz central frequency, 15 mm aperture, and 35 mm focal length was used,
which yields a lateral resolution of 800 µm and an axial resolution of 200 µm. For the
in vivo animal experiments, a focused high-frequency transducer (50 MHz, 3 mm
aperture, and 6 mm focal length) was used, which provides a lateral resolution of 61
µm on the focal point and an axial resolution of 15 µm.
For these experiments, laser pulses at 730 nm from the Ti:Sapphire laser were used
to image one 0.8 mm-diameter circular target (for cases 1–5) or three circular targets
(0.8 mm in diameter each) (for case 6) embedded in the background, which was a
scattering medium composed of a fat emulsion suspension (Intralipid) with India ink
added as an absorber. The method we used to determine the absorption and reduced
scattering coefficients of the phantoms in this study has been widely adopted in the
field of diffuse optical imaging and spectroscopy (Pogue and Patterson 1994, Hale
and Querry 1973, Madsen et al. 1992, van Staveren et al. 1991). The absorption of the
suspension is essentially due to water24–26, hence the ink provided a controlled way of
achieving a higher level of absorption (Madsen et al. 1992). It is also known that the
added ink should not have a significant impact on the scattering coefficient (Madsen
et al. 1992). The targets were transparent tube containing different concentrations
of Intralipid and India ink, which were used to simulate heterogeneities. For cases
1–3, the single target had an absorption coefficient of 0.07 mm–1 and was located at
a depth of 0.0, 4.8, and 7.6 mm below the surface, respectively. For case 4, the single
target with an absorption coefficient of 0.035 mm –1 was embedded at a depth of
2.1 mm. For case 5, the single target having an absorption coefficient of 0.021 mm–1
was placed at a depth of 1.1mm. For case 6, three targets with different absorption
coefficients (0.07 mm–1, 0.035 mm–1, and 0.021 mm–1, respectively) were embedded
at a depth of 1.9 mm. For all these cases, the absorption coefficient of the background
was 0.007 mm–1, and the reduced scattering coefficient of the background and tar-
gets were 1.0 mm–1.
5.4.1.3 Animal Experiments
Before the experiment of imaging the rat ear, one rat weighing 28 g was anesthetized
using urethane and the hairs of the ear were removed. The laser beam at 532 nm
generated by a Nd:YAG laser was split into two subbeams and coupled into two
optical fiber bundles. A focused ultrasound transducer (50 MHz, 3 mm aperture,
and 6 mm focal length) was used to receive the induced photoacoustic waves. A 2D
moving stage mounted with the optical fiber bundles and transducer carried out a
2D raster scanning. The raster scanning provided 200 A-lines along each direction
with an interval of 60 μm over a distance of 12 mm. For the experiment of imaging
the brain of a rat weighing 25 to 30 g, the brain area of interest was gently depilated,
and the scalp was removed before the imaging (Figure 5.12a). Pulsed light at 1064 nm
from the Nd:YAG laser and at 730 nm from the Ti:Sapphire laser were used for
multispectral imaging. B-scan photoacoustic signals were recorded by a focused
50 MHz high-frequency transducer. During the imaging, the rat was anesthetized
with urethane (1mg/g) and tightly positioned with a holder to avoid motion effect
due to seizure activity. The body temperature of the animal was maintained at 37°C
with a temperature controlling pad. At the end of study, all rats were sacrificed. All
procedures have been approved by the University of Florida Institutional Animal
Care and Use Committee (IACUC).
5.4.1.4 Image Reconstruction
5.4.1.4.1 Recovery of Absolute Optical Absorption Coefficient
Our PACM approach is based on a strategy that iteratively solves the radiative trans-
fer equation (RTE) and the photoacoustic wave equation (PWE) using finite element

method (FEM). In PACM, we first obtain the distribution of optical fluence, Φ(r )
through the finite element (FE) solution to the RTE from an initial (guess) distribu-

tion of optical absorption coefficient, 0a (r ) . For quantitative imaging of individual
blood vessels, the use of RTE is necessary and most rigorous to model the light prop-
agation in tissue since the absorption coefficient of blood vessels is high, making the
commonly used photon diffusion equation inappropriate for quantitative PAM. The
RTE used is described as follows:

( ⋅ + s + a )φ(r , ) = s
∫ S n −1
φ(r , )Θ( , ) d + q(r , ) (5.2)

where s isthe scattering coefficient; φ(r , ) is the radiance; q(r , ) is the source
term;
and ∈ S n−1 denotes a unit vector in the direction of interest. The kernel,
Θ( , ), is the scattering phase function describing the probability density that a
photon with an initial direction will have a direction after a scattering event.
We assume that the scattering phase function depends only on the angle between the
incoming and outgoing directions; thus
the commonly used Henyey-Greenstein scat-
tering function can be applied: Θ ( , ) =
Θ ( ⋅ ) = (1 − g 2 )/2π(1 + g 2 − 2 g cos γ ) ,
where γ is the angle between and , and −1 < g < 1. The optical fluence is

related to the radiance by Φ(r ) = ∫ S n −1 φ(r , )d .
Next, we computationally obtain the acoustic pressure, pkc ( x k , yk , z , t ) at each
transducer location along the surface of the medium/tissue (rk (xk , yk), k = 1,2,…M
transducer locations) using the following Helmholtz PWE (Yao and Jiang 2009):
2 1 ∂2 pkc (rk , z , t ) β a (rk , z )Φ(rk , z ) ∂ J (t )

pkc (rk , z , t ) − 2 2
=− , (5.3)
v0 ∂t Cp ∂t

where p is the pressure wave; v0 is the speed of acoustic wave in the medium; β is the
thermal expansion coefficient; Cp is the specific heat; and J (t ) = δ(t − t0 ) is assumed
in this study.
Finally, using the regularized Newton’s method, the matrix equation for inversion
of the absorption coefficient can be written as

(ℑ ℑ
T
k k + λI ) χ k = ℑTk ( pko − pkc ) , (5.4)
where pko and pkc are the observed and computed complex acoustic field data for k =
1,2,…M transducer locations; Δχk is the update vector for the absorption coefficient;
ℑ is the Jacobian matrix formed by ∂ pk /∂χ k at the boundary measurement sites; λ
is the regularization parameter determined by combined Marquardt and Tikhonov
regularization schemes (Jiang 2010, Jiang et al. 2006) (λ = 0.5 was used in this
study—we have found that the quality of inversion was not sensitive to the choice
of this parameter); and I is the identity matrix. Thus here an A-line image (χ k(z) or
μa,k(z)) is formed by updating absorption coefficient distribution via iterative solution
of Equations (5.2) to (5.4) so that an object function composed of a weighted sum of
the squared difference between computed and measured acoustic data can be mini-
mized. A B-scan or 2D image is obtained by combining all the A-line images and
a 3D image is provided by combining all the B-scan images. In these calculations,
the incident laser source strength and the scattering coefficient μs are estimated in
advance.
5.4.1.4.2 Recovery of Functional/Hemodynamic
Parameters (HbR, HbO2, BF and MRO2)
In most cases for microvasculature imaging, HbO2 and HbR are the two major
absorbers. Using multispectral measurements, the oxy- and deoxy-hemoglobin con-
centrations, [HbO2] and [HbR], can be obtained using the Beer-Lambert law:

(r , λ) = ε HbR (λ) [ HbR ](r ) + ε HbO2 (λ) [ HbO 2 ](r ) (5.5)
a
where ε HbR (λ) and ε HbO2 (λ) are the known molar extinction coefficients of HbR
and HbO2 at wavelength λ. Oxygen saturation can be calculated as sO2 = [HbO2]/
[HbT] where HbT is the total hemoglobin ([HbT] = [HbO2] + [HbR]).
We can further calculate the blood flow (BF) and rate of oxygen metabolism
(MRO2) in blood vessels from the functional parameters obtained above. The
method for such calculation is based on the principle of mass balance to the trans-
port of oxygen in a blood vessel segment (Tsai et al. 2003). To model oxygen trans-
port in a blood vessel by this principle, we consider a 1-D cylindrical vessel (blood
vessel) with Ri and Ro as the inner and outer radii, respectively, surrounded by the
biological tissues. In addition, we assume that all the oxygen (O2) diffusing out the
segment is consumed in a tissue region (Sharan et al. 2008). The law of mass con-
servation stipulates that the amount of O2 lost from a vascular segment must be
equal to the O2 flux diffused to the tissue, determined by the perivascular oxygen
gradients. The oxygen consumed by the tissues (organs) is supplied from three blood
vessels sources: capillaries, arterioles, and venules. If the oxygen supply of tissues
depends on the averaged oxygen saturation at the inlet and outlet of the tissues,
tissue oxygen saturation should represent the weighted average of the arterial and
venous saturation:
sO 2 = fsO 2,ti + (1 − f )sO 2,to (5.6)

where sO 2,ti and sO 2,to are the averaged oxygen saturation at the inlet (artery) and
outlet (vena) of the tissues, respectively. As such, mass balance for O2 in all tissues
(organs) based on global analysis yields the following estimation of intravascular
flux for a dynamic case, considering expressions of molar amount of oxy-hemoglobin
concentration of tissue and tissue oxygen saturation:
d OC BF sO 2,ti − sO 2 d [ HbT ] sO 2
sO 2 = − + [HbT]blood − , (5.7)
dt 4V [HbT] V [HbT] 1− f dt [HbT]

where BF is the mean blood flow for all the blood vessels inside the tissues and is
specified as the mean blood flow of tissues, OC = MRO2 is the mean oxygen con-
sumption for the whole tissue, [HbT] is the mean total hemoglobin concentration in
the tissues, [HbT]blood is the mean total hemoglobin concentration in the blood circu-
lating through the tissues, and V is the tissue volume and is assumed constant here. In
our calculations, we assumed [HbT]blood = 0.72 mM, f = 0.2, and sO2,ti = 0.98. Further
explanations on the determination of these parameters can be found elsewhere (Carp
et al. 2008).
Equation (5.7) is the mathematical model we have developed for predicting BF and
MRO2 from [HbT] and SO2. It is an ordinary partial differential equation that can be
solved iteratively by Runge-Kutta fourth-order method coupled with the FEM (Press
et al. 1992). Thus, mean BF and MRO2 can be recovered by fitting Equation (5.7) to
time-resolved tissue oxygenation measurements. The fitting method is described as
follows: with any given initial values for OC and BF within the specified range, this
scheme is to optimize the OC and BF parameters based on the solution of Equation
(5.7) to reach the minimized objective function in Equation (5.8):
∑(sO
2
Min: F = m
2i − sOc2i ) (5.8)
i =1
in which sO 2mi is the measured oxygenation parameter from M discrete time points,
and sOc2i is the oxygenation parameter calculated from Equation (5.6) for the same
M time points. Note that the BF and OC are assumed constant during the measure-
ments for the specified time range, due to the need for a sufficient time interval to
obtain stable fitting results.
5.4.2 Results
The PACM reconstruction algorithm (described above) is first validated using
six sets of tissue-mimicking phantoms with known optical properties (Table 5.2).
TABLE 5.2
Exact and Recovered Size and Absorption Coefficient Value
of Target for the Phantom Experiments
µa of Target (mm–1) Size of Target (mm)
Depth of the
Case # Target (mm) Exact Recovered Exact Recovered
I 0 0.071 0.90
II 4.8 0.07 0.071 0.90
III 7.6 0.072 0.95
IV 2.1 0.035 0.034 0.95
V 1.1 0.021 0.020 0.8 1.0
VI 1.9 0.07 0.068 0.9
0.035 0.033 1.0
0.021 0.019 1.1
Figures 5.10a to 5.10f show the reconstructed absorption coefficient images from
these phantom experiments. We see that the object(s) in each case are clearly
detected by our PACM approach. In particular, these images are quantitatively accu-
rate in terms of the absorption coefficient of target. This conclusion is further con-
firmed by the quantitative absorption coefficient profiles for these images provided
in Figures 5.10g to 5.10h. We found that the relative error of the recovered absorption
coefficient of target compared to the exact value is less than 4.7% for the single-target
cases 1 to 5, while it ranges from 2.8 to 9.5% for the multitarget case 6 (Table 5.2).
We note that the target having higher contrast (cases 1–3) is better reconstructed than
for the lower contrast cases (case 4).
A rat ear experiment was chosen to further validate the algorithm and show its
possible application to functional imaging based on physiologically specific endog-
enous optical absorption contrasts in biological tissue. In Figure 5.11a we present the
maximum amplitude projection (MAP) images projected along the vertical direction
(z-axis) to the orthogonal plane where seven orders of vessel branching, indicated
by numbers 1 to 7, can be observed in the image. The B-scan image and the recon-
structed absorption coefficient image obtained by our FE-based PACM algorithm
are shown in Figures 5.11b and 5.11c, respectively. These images are in the vertical
plane (x-z plane) at the location indicated by the dashed line in Figure 5.11a. It can
be clearly seen that all of the vessel branching can be reconstructed using both of the
methods. The laser applied in this experiment was at a wavelength of 532 nm, and
from Figure 5.11c we found that at most locations within the blood vessel, the value
of the reconstructed absorption coefficient is in the range of 20~30 mm–1, which is
in good agreement with that of blood vessels at 532 nm reported in the literature
(Barton et al. 1998, Kienle et al. 1996).
(mm–1) (mm–1) (mm–1) (mm–1) (mm–1)
3 1 2
0.07 –3 0.07 –5 0.07 0.0.35
2 0 1 0.02
0.06 0.06 –6 0.06 0.03
1 –4 –1 0
0.05 0.05 –7 0.05 0.025
0 –5 0.04 –2 0.015
0.04 0.04 0.02 –1
0.03 –6 0.03 –8 0.03 –3
–1 y 0.015 –2
0.02 –7 0.02 –9 0.02 0.01
–2 1 mm –4 0.01 –3
0.01 0.01 –10 0.01
z –8 0.005
0 2 4 0 2 4 0 2 4 0 2 4 0 2 4
(a) (b) (c) (d) (e)
(mm–1)
1 0.100 0.100
Case 1 Case 6
0 0.06 Case 2
0.075 Case 3 0.075
0.05 Case 4
–1
0.050 Case 5 0.050
0.04
–2
0.03 0.025 0.025
–3
0.02
–4 0.000 0.000
0.01 0 1 2 3 4 5 6 0 4 8 12

0 2 4 6 8 10 12 14 16 Y Position (mm) Y Position (mm)

(f ) (g) (h)
FIGURE 5.10 Reconstructed absorption coefficient images from the phantom experiments. (a) Case 1. (b) Case 2. (c) Case 3. (d) Case 4. (e) Case 5.
Photoacoustic Microscopy and Photoacoustic Computed Microscopy
(f) Case 6. (g) and (h) Quantitative plots: reconstructed absorption coefficient profiles along y = 0.0 mm for case 1, y = –4.8 mm for case 2, y = –7.6 mm
for case 3, y = –2.1 mm for case 4, y = –1.1 mm for case 5, and y = –1.9 mm for case 6.
147
1 2 3 4 5 6 7 5 7
6
3
4
1 2
1 mm
(b)
(mm–1)
5 7 60
6
4 40
2 3
1 20
1 mm
(a) (c)
FIGURE 5.11 In vivo imaging of the blood vessels in a rat ear. (a) The MAP image of the
photoacoustic signals projected on the orthogonal plane, with seven small vessels that can be
observed in the image as indicated by numbers 1–7; (b) B-scan image in the vertical plane at
the location indicated by the dashed line in (a); (c) reconstructed absorption coefficient image
in the vertical plane at the location indicated by the dashed line in (a).
We then demonstrate the ability of PACM in in vivo imaging HbO2, HbR, sO2,
cerebral blood flow (CBF), and cerebral rate of oxygen metabolism (CMRO2) using a
rodent model in nondiseased condition. Figure 5.12b shows a typical in vivo B-scan
PAM image of a single blood vessel located 1.5 mm below the surface of the rat
brain with intact skull (along y-z plane) at a selected time point, which was taken at
the location indicated by the dashed line (Figure 5.12a) (a total of 10 B-scan images
were obtained over 60 s duration). Electroencephalographic (EEG) signal from one
electrode is given in Figure 5.12c. For simplicity, the reconstructions of quantitative
absorption coefficient and all the hemodynamic response parameters were conducted
over the square area shown in Figure 5.12b. Using the PACM reconstruction algo-
rithm, we obtained the absorption coefficient images at 730 and 1064 nm over 60 s
duration, and Figures 5.13a to 5.13h show the absorption coefficient images for the two
wavelengths at four selected time points, t1 = 12s, t2 = 24s, t3 = 42s, and t4 = 54s. These
high-resolution images not only allow for the accurate recovery of the size/shape/
location of single blood vessel, but the absolute absorption coefficient itself (indicated
by the gray scale on right). The absorption coefficient images at 730 and 1064 nm
allowed us to obtain HbO2 (Figures 5.14a–5.14d), HbR (Figures 5.14e–5.14h), and
sO2 (Figures 5.14i–5.14l) using Equation (5.5) at the four selected time points as well
as at other time points. It can be seen that these functional parameters can also be
reconstructed quantitatively over time. Based on these results, the reconstructed CBF
and CMRO2 images at four different times are provided in Figures 5.15a–5.15d and
5.15e–5.15h. To closely observe dynamic evolvement of the functional parameters
including HbO2, HbR, sO2, CBF, and CMRO2, we calculated the average value of
x 0.5
–0.3 Surface
y –0.6
0.4
–0.9 FE reconstruction
area
–1.2
0.3
Z (mm)
–1.5
–1.8 0.2
–2.1
–2.4 0.1
–2.7
–3.0 0
0.3 0.6 0.9 1.2 1.5
1 mm
Y (mm)
(a) (b)
0.04
EEG Intensity (mV)
0.02
0
–0.02
–0.04
10 20 30 40 50 60
t (s)
(c)
FIGURE 5.12 (a) Photograph of the rat brain with intact skull under PAM imaging. The
arrow/dashed line indicates the B-scan path for PAM. (b) Typical B-scan PAM image. Note
that the blood vessel imaged is ~1.5mm below the surface. (c) EEG signal recorded at one
electrode close to the scanned region.
these parameters over 60 s time course in a region of interest (ROI), depicted by

a 0.1 mm-diameter circle shown in Figure 5.14b, and the results are presented in
Figure 5.16 where the EEG signal is also given for comparison.
5.4.3 Discussion
The reconstructed results of optical absorption coefficient obtained from the phan-
tom experiments and the rat ear are both qualitatively and quantitatively in terms of
the location, size, shape, and optical property values of the target and background,
which indicates that the PACM method has the capability to provide absolute func-
tional imaging based on physiologically specific endogenous optical absorption coef-
ficient. This capability is confirmed with an animal experiment by using a rodent
model (see Section 5.4.2), but in principle any animal model associated with neural
activation/stimulation could be used.
(mm–1) (mm–1)
0 0
–0.1 0.08 –0.1 0.08
–0.2 –0.2
0.06 0.06
–0.3 –0.3
y
–0.4 0.04 –0.4 0.04
–0.5 0.1 mm –0.5
z 0.02 0.02
0 0.2 0.4 0.6 0 0.2 0.4 0.6
(a) (b)
(mm–1) (mm–1)
0 0
0.08
–0.1 –0.1
0.08
–0.2 0.06 –0.2
0.06
–0.3 –0.3
0.04 –0.4
–0.4 0.04
–0.5 –0.5
0.02 0.02
0 0.2 0.4 0.6 0 0.2 0.4 0.6
(c) (d)
(mm–1) (mm–1)
0 0
0.08
–0.1 –0.1 0.08
–0.2 0.06 –0.2 0.06

–0.3 –0.3
0.04 0.04
–0.4 –0.4
–0.5 –0.5
0.02 0.02
0 0.2 0.4 0.6 0 0.2 0.4 0.6
(e) (f )
(mm–1) (mm–1)
0 0
0.08
0.08
–0.1 –0.1
–0.2 0.06 –0.2 0.06
–0.3 –0.3
0.04 0.04
–0.4 –0.4
–0.5 –0.5
0.02 0.02
0 0.2 0.4 0.6 0 0.2 0.4 0.6
(g) (h)
FIGURE 5.13 Reconstructed absorption coefficient images using the in vivo rodent model.
(a)–(d): 730 nm; (e)–(h): 1064 nm at four different time points (a and e: t = 12 s; b and f: t =
24 s; c and g: t = 42 s; d and h: t = 54 s).
( g/L) ( g/L) ( g/L)

0 40 0 0 40
40
–0.1 –0.1 –0.1
–0.2 30 –0.2 –0.2 30
30
–0.3 –0.3 –0.3
20 20 20
–0.4 y –0.4 ROI –0.4
–0.5 0.1 mm –0.5 –0.5
z 10 10 10
0 0.2 0.4 0.6 0 0.2 0.4 0.6 0 0.2 0.4 0.6
(a) (b) (c)
( g/L) ( g/L) ( g/L)

0 40 0 10 0 8
–0.1 –0.1 –0.1
8
–0.2 30 –0.2 –0.2 6
–0.3 –0.3 6 –0.3
20 4
–0.4 –0.4 4 –0.4
–0.5 –0.5 –0.5 2
10
2
0 0.2 0.4 0.6 0 0.2 0.4 0.6 0 0.2 0.4 0.6
(d) (e) (f )
( g/L) ( g/L)
0 0 0
–0.1 6 –0.1 8 –0.1
0.6
–0.2 –0.2 –0.2
–0.3 –0.3 6 –0.3
4 0.4
–0.4 –0.4 4 –0.4
–0.5 2 –0.5 –0.5 0.2
2
0 0.2 0.4 0.6 0 0.2 0.4 0.6 0 0.2 0.4 0.6
(g) (h) (i)
0 0.8 0 0.8 0 0.8

–0.1 –0.1 –0.1
–0.2 0.6 –0.2 0.6 –0.2 0.6
–0.3 –0.3 –0.3 0.4

0.4 0.4
–0.4 –0.4 –0.4
–0.5 0.2 –0.5 0.2 –0.5 0.2
0 0.2 0.4 0.6 0 0.2 0.4 0.6 0 0.2 0.4 0.6
(j) (k) (l)
FIGURE 5.14 Reconstructed functional images using the in vivo rodent model. (a)–(d):
HbO2, (e)–(h): 1 HbR, and (i)–(l): sO2 at four different time points (a, e, and i: t = 12 s; b, f,
and j: t = 24 s; c, g, and k: t = 42 s; d, h, and l: t = 54 s).
In this study we first obtained the reconstructed absorption coefficient images at

two wavelengths (730 and 1064 nm) at four different points (12 s, 24 s, 42 s, and 54
s) as shown in Figure 5.13. Functional information including HbO2, HbR, and sO2 at
these time points is presented in Figure 5.14. We note that the total hemoglobin con-
centration ([HbT] = [HbO2] + [HbR]) of the blood vessel is in the range of 40 to 50 g/l,
which agrees well with that described in the literature (Berkow et al. 1997). The
dynamic changes of these functional parameters in the ROI are also clearly notable
in Figures 5.16b to 5.16d. Inspecting the filtered EEG signal over time (Figure 5.16a),
ml/ml/s ml/ml/s
× 10–3 × 10–3
0 0
8
–0.1 –0.1 8
–0.2 6 –0.2
6
–0.3 –0.3
–0.4 y 4 –0.4 4
–0.5 0.1 mm 2 –0.5
z 2
0 0.2 0.4 0.6 0 0.2 0.4 0.6
(a) (b)
ml/ml/s ml/ml/s
× 10–3 × 10–3
0 0
–0.1 8 –0.1 8
–0.2 –0.2
6 6
–0.3 –0.3
–0.4 4 –0.4 4
–0.5 –0.5
2 2
0 0.2 0.4 0.6 0 0.2 0.4 0.6
(c) (d)
µmol/ml/s µmol/ml/s
× 10–3 × 10–3
0 0
8 8
–0.1 –0.1
–0.2 6 –0.2
6
–0.3 –0.3
–0.4 4 –0.4 4
–0.5 2 –0.5
2
0 0.2 0.4 0.6 0 0.2 0.4 0.6
(e) (f )
µmol/ml/s µmol/ml/s
× 10–3 × 10–3
0 0
–0.1 –0.1 8
8
–0.2 –0.2
6 6
–0.3 –0.3
–0.4 4 –0.4 4
–0.5 –0.5
2 2
0 0.2 0.4 0.6 0 0.2 0.4 0.6
(g) (h)
FIGURE 5.15 Reconstructed CBF (a)–(d), and CMRO2 (e)–(h) images at t = 12 s (a,e), 24 s
(b,f), t = 42 s (c,g) and t = 54 s (d,h). In the calculation, the initial parameters used were
[HbT]blood = 0.72 mM, f = 0.2, and sO2,ti = 0.98 (see Section 5.4.1.4.2 for the explanation of
these initial parameters). In addition, due to the high nonlinear distribution of sO2 over time,
the sO2 distribution curve was first separated into several approximated linear segments to
improve the fitting accuracy of CBF and CMRO2. The mean CBF and CMRO2 were fitted for
each linear segment based on different initial values of [HbT] and sO2.
EEQ Intensity (mV)

0.02 (a)
0.01
0
–0.01
–0.02
10 20 30 40 50 60
40
(b)
HbO2 (g/l)
36
32
28
0 10 20 30 40 50 60
t (s)
11 (c)
HbR (g/l)
9
7
5
0 10 20 30 40 50 60
t (s)
0.88
(d)
0.84
sO2
0.8
0.76
0 10 20 30 40 50 60
t (s)
0.009
(e)
CBF (ml/ml/s)
0.0086
0.0082
0.0078
0 10 20 30 40 50 60
t (s)
0.0092 (f )
CMRO2
0.0088
0.0084
0.008
0 10 20 30 40 50 60
FIGURE 5.16 Comparison of the EEG signal and the average values of the recovered func-
tional parameters of the ROI during the entire time course. (a) Filtered EEG signals; (b) aver-
age HbO2 of the ROI; (c) average HbR of the ROI; (d) average sO2 of the ROI; (e) average CBF
of the ROI; (f) average CMRO2 of the ROI.
we find that two significant spikes at 20 and 50 s in the EEG signals can also be
identified in Figures 5.16b and 5.16c.
Using Equations (5.7) and (5.8), we have also calculated CBF and CMRO2 over
the total time course. We can observe significant changes in CBF and CMRO2 in
terms of the shape/size of blood vessel and values of CBF/CMRO2. We also see
from the peak values of CBF shown in Figure 5.15 that the recovered blood flow
values (7.5–8.5 ml/ml/s, or 45–51 ml/100 ml/min) are in good agreement with the
reported CBF of rats (10–120 ml/100 ml/min) and of humans (20–160 ml/100 ml/
min) (Hernandez et al. 1978, Sharples et al. 1995). Average values of CBF and
CMRO2 in the ROI are provided in Figures 5.16e and 5.16f. Interestingly, CBF and
CMRO2 spikes stronger in intensity than that seen in HbR and HbO2 are clearly
notable.
While it is a novel approach to monitor the dynamic response of functional
parameters quantitatively, PACM presented here has some limitations and needs to
be improved in the future. In this study, we estimated a homogenous distribution of
scattering coefficient in advance when the absorption coefficient was recovered. For
the animal experiments the reduced scattering coefficient was assumed as 1.0/mm.
It is well known that the propagation of light in turbid media is strongly character-
ized by optical absorption and scattering; thus this approximation would introduce
a negative impact on the accuracy of the recovery of absorption coefficient. This
impact, however, was small since the contribution of scattering to the photoacoustic
data is much less than that of absorption, if the estimated distribution of scattering
coefficient is in a reasonable range. For example, based on our numerical simula-
tions, when the estimated error of the reduced scattering coefficient is within the
range of ±50% of the actual value, the error of the recovered absorption coefficient
is less than 5%. Recovering absorption and scattering coefficients simultaneously
will improve the accuracy of the reconstructed results, but it will suffer from strong
nonuniqueness for such inversion of multiple parameters. In this study, since we
focus on introducing a new method exploiting the recovery of absorption coefficient
and other associated parameters of biological tissue, this approximation for scatter-
ing coefficient was used for simplicity.
To produce the functional parameters in the blood vessel by PACM, laser beams
at 730 and 1064 nm were used in this study due to the limitation of our current hard-
ware. The contribution of water to the absorption coefficient should be considered,
since the molar extinction coefficient of water at the wavelength 1064 nm cannot be
ignored compared to that of the oxygenated and deoxygenated hemoglobin. In this
study, we made an assumption that the ratio of the concentration of water to the total
hemoglobin in blood was about 55%:45%, based on the fact that the blood plasma
makes up about 55% of total blood volume. Dynamic images (i.e., CBF and CMRO2)
with similar values as the ones obtained under the assumption were obtained if dif-
ferent ratios other than 55% were used in the calculations. The impact of water can
be better considered by using laser beams at three different wavelengths, or at two
wavelengths (for example, 730 and 850 nm), at which the molar extinction coefficient
of water is relatively low.
The model we used to calculate CBF and CMRO2 in blood vessels from the func-
tional parameters is based on the principle of mass balance to the transport of oxygen
in a blood vessel segment. Some assumptions are made in order to use this model.
For example, oxygen consumption and blood flow are assumed constant during the
measurements, due to the need for a sufficient time interval to obtain stable fitting
results. While this is a very good assumption for oxygen consumption, blood flow
may change during tissue relaxation, indicating that the fitted blood flow value rep-
resents a measure of an average blood flow during the measurement period. The
accuracy of the fitting results can be improved by increasing the temporal resolution
of our imaging system.
In our calculations we assumed [HbT]blood = 0.72 mM, f = 0.2, sO2,ti = 0.98. The
values of sO2,ti and f were taken from the literature (An and Lin 2002, Duong and
Kim 2000). With respect to the choice of [HbT]blood, the normal hemoglobin concen-
tration is in the range of 120–160 g/L for females (Berkow et al. 1997), and since
the hematocrit values of the capillary vessels is known to be two to five times lower
than the systemic hematocrit (Desjardins and Duling 1987). For the purposes of this
study, we assumed a value of 140 g/L for systemic hemoglobin concentration and a
three-times dilution factor in the small blood vessels, resulting in a value of 0.72 mM
for [HbT]blood. We note that the choice of [HbT]blood does influence the flow results,
but it does not affect the calculated oxygen consumption, because the inflow and out-
flow terms appear as part of a product with [HbT]blood in the mass balance equations.
In summary, we have demonstrated that PACM is capable of in vivo imaging
a full set of functional parameters at the small vessel level. The dynamic changes
are comparable with concurrently recorded EEG signals. The spatial resolution of
PACM is the same as PAM, scalable with the central frequency of the focused trans-
ducer, i.e., higher resolution at the single cell level is achievable with the sacrifice of
penetration depth. The imaging speed of the current PACM is relatively slow (6 s), but
it can be improved considerably by using faster lasers. We expect PACM will be a
valuable tool for neuroscience research where hemodynamics associated with micro-
vasculature in response to neural activation/stimulation need to be imaged. PACM
will also be applicable to visualize microvasculature dynamics involved in tumor
angiogenesis and in inflammatory joint diseases.
5.5 PHOTOACOUSTIC MICROSCOPY BASED ON

ACOUSTIC LENS WITH VARIABLE FOCAL LENGTH
For most ARPAM implementations, axial scanning of the detector is required owing
to the degrading resolution out of the acoustic focus. However, in an example of endo-
scopic imaging configuration, one issue comes up that little room could be spared for
the axial scanning of the detector. Thus, ARPAM cannot be performed optimally in
this case. Here we present an acoustic lens with variable focal length to eliminate
the axial scanning of the detector, and to make it possible to miniaturize ARPAM
systems for important clinical applications such as endoscopic examination.
Acoustic lenses have been commonly used in photoacoustic imaging systems for
either focusing (Niederhauser et al. 2004) or diverging acoustic wave (Xia et al. 2011).
However, these acoustic lenses are solid based, which brings up the issues of high
acoustic impedance and low transmission efficiency. In particular, these solid-based
acoustic lenses have a fixed focal length once the lens is fabricated. Recently the field
of optofluidics has emerged to enlighten the idea that solid-based optical components
could be replaced by liquids or fluids with better performance and novel functional-
ities (Psaltis et al. 2006). Optical liquid or fluidic lenses have been demonstrated to
have an adjustable focal length by either pneumatically control (Zhang et al. 2004) or
flow rate management (Song et al. 2009, 2010). We have described a liquid acoustic
diverging lens that can enlarge the acceptance angle for PAT in Chapter 3. Here, we
demonstrate a liquid acoustic lens with an adjustable focal length and apply it to PAM.
The tunable function of this liquid acoustic lens is realized by pneumatically
controlling the infusion volume of the liquid. Specifically, a syringe pump (KDS
210, KD Scientific) is used to accurately infuse or withdraw liquid into or from the
lens chamber. In this way, the lens interface, which is an elastic membrane, can be
tuned with its curvature. Therefore, the focal length of this acoustic lens can be
adjusted simply by changing the infusion volume of the liquid. The relationship
between the curvature of the interface and the infusion volume of the liquid is char-
acterized and shown in Figure 5.17. Assuming that the lens interface has a spherical
shape, the infusion volume can mathematically determine the radius of the inter-
face curvature (solid line in Figure 5.17) via V = πh2(r–h/3), where V represents the
infusion volume of the liquid and h and r denote the thickness of the lens and radius
of interface, respectively. The experimental results show a good agreement with the
mathematical prediction.
Setting this liquid acoustic lens on its focusing or diverging mode can be real-
ized by purposely choosing the working liquid inside the lens chamber, which by
principle is similar to manipulating optofluidic lens (Song et al. 2011). Here, we
choose silicone oil as the working liquid, which has a relative refractive index of 1.4
0.2
0.18 Theoretical curve
Experimental results
Curvature of the Interface (1/mm)
0.16
0.14
0.12
0.1
0.08
0.06
0.04
0.02
0
0 50 100 150 200 250 300 350
Infusion Volume (µL)
FIGURE 5.17 Curvature of the lens interface as a function of the infusion volume of liquid.
The inset photographs illustrate the lens interface tuned by the infusion volume of the liquid.
Without lens Infusion 30 µL Infusion 60 µL Infusion 90 µL Infusion 120 µL
1 mm
(a)
26
Fitting curve
24 Experimental results
Focal Length (mm)
22
20
18
16
14
50 100 150 200 250 300
Infusion Volume (µL)
(b)
FIGURE 5.18 (a) Pressure distribution mappings of the transducer with and without the liq-
uid acoustic lens. The arrow shows the propagation direction of the acoustic beam; (b) focal
length of the liquid acoustic lens as a function of the infusion volume of liquid.
and density of 0.986 g/cm3. The higher refractive index (relative to water) as well
as the convex shape of the lens will enable an ability of focusing acoustic beam.
To characterize the focal length of this liquid acoustic lens, a transducer (30 MHz,
Olympus) was used in its transmission mode for providing ultrasound pulses. The
liquid acoustic lens was attached to the transducer head to focus the acoustic wave.
A hydrophone (HGL-0200, ONDA) was used to scan laterally and axially while
detecting the ultrasound pulses from the transducer (Song et al. 2013).
Figure 5.18a shows the comparison of pressure distributions between the trans-
ducers with and without the liquid acoustic lens. The arrow in the figure shows the
propagation direction of the acoustic beam. The acoustic wave emitted from the
transducer has a slightly focused wavefront (refer to the pressure mapping of trans-
ducer without the lens). When the transducer is working with the lens attached, the
acoustic wave can be focused with variable focal length. Since the infusion volume
of liquid determines the curvature of the lens, increasing the infusion volume by
Holder Transducer Tumors buried at

different depths
2
1
Liquid acoustic lens
Hairs buried at
Laser pulsed illumination different depths
1 2 3
Translational stage
Transparent glass slide
FIGURE 5.19 Schematic of the experimental setup. Two groups of tests were carried out to
validate the functionality of the adjustable liquid acoustic lens: (1) Three hairs were buried
at different depths within the background phantom with 4mm distance between each hair;
(2) two tumor samples were buried at different depths within the background phantom with a
distance of 1.5 cm between each.
controlling the syringe pump can lead to a shortening of the focal length. Figure 5.18b
shows the focal length as a function of the infusion volume of the liquid. This test
gives a validation that the liquid acoustic lens can effectively focus acoustic wave
with adjustable focal length, and provides a calibration of relationship between the
focal positions and the infusion volume.
The advantage of this liquid acoustic lens with an adjustable focal length lies in
imaging objects at different depths without axial scanning of the detector. We designed
and carried out two groups of tests to demonstrate this advantage: one using human
hair as microscale targets, and the other using excised tumors from a mouse as mac-
roscale targets (Figure 5.19). The targets were buried in a background with an absorp-
tion coefficient of μa= 0.007 mm–1 and a reduced scattering coefficient μ′s = 1 mm–1.
A 30 MHz transducer (Olympus) was used to collect the signal from the hair targets,
and a 10 MHz transducer (V315, Panametrics) was employed for the tumor targets.
The targets were illuminated and scanned by laser pulses with wavelength 532 nm and
pulse duration 6 ns (NL 303HT, EKSPLA) coupled with a translational stage (Zaber
Tech T-LSM200A). Upon the energy deposition, an acoustic wave was generated from
the local thermo-expansion and reflected by a glass slide to the ultrasound detector.
The liquid acoustic lens was attached to the transducer and mounted on a hard-resin
holder. The liquid was injected into the lens chamber through needle and tubing by a
syringe pump (KDS 210, KD Scientific). During the data acquisition, the axial distance
between the targets and the transducer was fixed, and the translational stage provided a
lateral scanning under each condition of infusion volume of the liquid.
For the hair imaging, three hairs were aligned in parallel and buried at different
depths within the background phantom with 4 mm distance between each other. The
Without lens Infusion Infusion Infusion

10 uL 30 uL 100 uL
Hair 1
1 mm
Hair 2
Hair 3
FIGURE 5.20 Images of hair by photoacoustic microscopy. Images of the three hairs in
left column were captured by the 30 MHz transducer without the liquid acoustic lens, and
the other three columns of images were captured with the lens under different conditions of
infusion volumes.
photoacoustic signal was firstly collected without using the liquid acoustic lens and
processed with maximum amplitude projection (MAP) method. The images of the
hairs (left column in Figure 5.20) show a broadened profile regarding the thickness
of the hair. This lack of image fidelity is owing to the finite size of the effective
diameter of the transducer. Considering the transducer has an effective diameter
of several millimeters, the hairs were imaged to have approximately the same size
regarding the thickness. After the transducer was equipped with the liquid acoustic
lens, the PAM system managed to obtain sharper images of the hairs (Figure 5.20)
when the focus of the lens was axially adjusted to coincide with the position of the
hair. During the test, the focal length was continuously shortened by increasing the
infusion volume of liquid. Thus, hair 3, which was the farthest from the transducer,
was firstly focused and imaged under a condition of low infusion volume of 10 µL.
With the infusion volume going higher and focal length getting shorter, hair 2 and
hair 1 came to their focus afterward and were sharply imaged under the infusion
conditions of 30 µL and 100 µL, respectively.
For the tumor imaging, two tumor samples with an approximately meniscus shape
(see the photograph in Figure 5.21) were buried in the background with a distance of
1.5 cm between each other. The superficially buried tumor 1 was located 0.5 and 5 cm
away from the surface and transducer, respectively. In this test, a 10MHz transducer
was used to collect the photoacoustic signal. A liquid acoustic lens was attached to
the transducer with its focal length varied by adjusting the infusion volume of the liq-
uid. When the infusion volume went to 50 µL, tumor 2, which was farther away from
the transducer, came into focus. Under this condition, tumor 2 was imaged to have
a meniscus shape that was approximately the same as its real shape, while tumor
Tumor 1 Tumor 1 Tumor 1
Infusion 50 uL 5 mm Infusion 150 uL Infusion 250 uL
Tumor 2 Tumor 2 Tumor 2
Infusion 50 uL Infusion 150 uL Infusion 250 uL

(a) Photoacoustic microscopy using liquid acoustic lens
2
1
(b) Photograph of the tumor samples
FIGURE 5.21 (a) Images of the tumors captured by a 10 MHz flat-head transducer with
liquid acoustic lens attached under different conditions of infusion volumes; (b) photograph
of the tumor samples.
1 was out of focus and had a blurred image (Figure 5.21a). As the infusion volume
was increased to 250 µL leading to a shorter focal length, tumor 1, which was closer
to the transducer, came into focus and was sharply imaged. Figure 5.21b shows the
photograph of the tumor samples for reference. Using the liquid acoustic lens, two
targets with an axial distance of 1.5 cm can be imaged without the implementation
of axial scanning but alternatively by adjusting the focal length of the acoustic lens.
5.6 CONFOCAL PHOTOACOUSTIC MICROSCOPY

USING A SINGLE MULTIFUNCTIONAL LENS
To enhance the sensitivity of PAM, acoustic and optical confocal is commonly used
with separate acoustic and optical converging strategies and complicated system con-
figurations. For ARPAM, a conical lens and an optical condenser are commonly used
to weakly focus the illumination beam, and complicated adjustments are required to
coaxially overlap the optical focus with the ultrasonic focus. ORPAM systems usually
employ both acoustic lens and objective lens to achieve acoustic and optical focus,
and a special combiner that reflects/transmits acoustic waves and transmits/reflects
optical beams to realize coaxially confocal of optical illumination and acoustic detec-
tion (Yuan et al. 2012). Both configurations make the PAM systems complicated and
experience based, especially for ORPAM systems. We have described a liquid variable-
focus lens to effectively achieve either acoustic diverging or converging, and evaluated
it using both PAT imaging system and ARPAM imaging system (see Sections 3.8 and
5.5). This liquid lens has several advantages: (1) it has inherent low acoustic impedance
that can effectively reduce the reflective loss of acoustic signals; (2) it has tunable focal
length that enables axial adjustment of acoustic focus without mechanical scanning;
and (3) it is cheap and easily replaceable. A liquid-filled optical lens has been already
widely used and offers the potential to minimize and simplify the optical systems
(Knollman et al. 1970, Sugiura and Morita 1993). In principle, the liquid-filled acoustic
and optical lens performs similar physical functions based on the difference of acoustic
or optical refractive index between the internal and external liquids. Here, we propose
the use of two different types of liquids but with the same relative acoustic and optical
refractive index to achieve coaxial confocal of optical illumination and acoustic detec-
tion. We use the new liquid lens to build a reflection-mode ORPAM and evaluate its
performance with phantom and in vivo animal experiments.
Similar to most ORPAM imaging systems, as shown in Figure 5.22a, the visible
laser beam (532 nm) generated by a Nd:YAG pulsed laser (Surelite I-20, Continuum,
CA) was firstly attenuated, then coupled into a single-mode fiber (P1-460B-FC-1,
Thorlabs) using two neutral density filters and one objective lens. The whole imag-
ing probe consisting of an optical collimator (F260FC-A, Thorlabs), a ring-shaped
5 MHz transducer (external diameter: 11 mm, internal diameter: 5 mm) and a liquid
lens with size of 11.5 mm in diameter was mounted on a 2D scanner. The captured
photoacoustic signals were amplified (5072PR, Olympus), digitalized (NI5124,
National Instruments), and stored in the computer. A beam splitter and an ultrafast
photodiode were used to record the pulse fluctuation that was used to calibrate the
imaging data. Figure 5.22b is the close view of the imaging probe shown in the
dashed rectangle in Figure 5.22a. The laser beam was firstly collimated by the opti-
cal collimator with an output size of 2.8 mm in diameter, then focused by the liquid
lens infused with the cinnamon oil through needle and tubing by a syringe pump
(KDS 210, KD Scientific). The ring-shaped transducer was confocally positioned
with the optical illumination by the same lens. The tank was filled with a mixture
of glycerol and water.
Under the condition of fully infused biconvex lens, both the optical and acoustic
focal length can be formulated as follows (Song et al. 2009, 2010, 2013):
r
f= . (5.9)
2(n − 1)
where r is the radius of the interface curvature, and n is the optical/acoustic relative
refractive index which is calculated by the following equations:
ninternal
noptical = (5.10)
nexternal
Vexternal
nacoustic = (5.11)
Vinternal
ND BP Lens
SMF
Nd: YAG
pulsed laser
PD
2D scanner
PC
Pre-amplifier
(a)
Acoustic beam
Collimator
Optical beam
Ring-shape
transducer
Liquid lens
Cinnamon
oil Needle
Glycerol and
water mixture
(b)
FIGURE 5.22 Schematic of the ORPAM imaging system (a) and imaging probe (b). ND:
neutral density filters, PD: photodiode, BP: beam splitter, SMF: single mode fiber.
where ninternal and nexternal represent the optical refractive index of internal and
external liquids, and Vinternal and Vexternal are the sound velocity of internal and exter-
nal liquids. Coaxial confocal of optical illumination and acoustic detection can be
achieved as long as the following conditions are satisfied: (1) the internal liquid has
higher refractive index and lower sound velocity compared with the external liquid
and (2) both the internal and external liquids are optical and acoustic transparent
and nontoxic. Cinnamon oil with impressively high optical refractive index (1.63)
and modest sound velocity (1480 m/s) serves well as the internal liquid in the area of
optical liquid lens and microfluidics. We already described the use of pure glycerol
with higher sound velocity (1920 m/s) for the fabrication of an acoustic diverging
lens (Section 3.8). We note that the pure glycerol has a lower optical refractive index
(1.47) compared with that of cinnamon oil. Hence, we can mix glycerol with water
to make the acoustic and optical relative refractive index identical.
Acoustic Optical
1.3
1.25
Relative Refractive Index
1.2
1.15
1.1
1.05
1
0 0.2 0.4 0.6 0.8 1
Water Percentage by Weight
(a)
Optical ray tracing Acoustic pressure distribution
Lens
2 mm
19 mm 19 mm Max
Min
(b)
FIGURE 5.23 (a) Acoustic and optical relative refractive index versus weight percentage of
water in the mixture. (b) Optical rays and acoustic pressure distribution when the lens was
flat and fully infused.
In Figure 5.23a, we plotted the curves of acoustic (triangles) and optical (squares)

relative refractive index versus the weight percentage of water in the mixture. The
acoustic and optical relative refractive index decreased and increased, respectively,
when the weight percentage of water was increased. Both the optical and acoustic
relative refractive index was 1.15 when the weight percentage of the water was 42%.
Given that the radius of lens and acoustic transducer was 5.5 mm and the liquid lens
was fully infused, the numerical aperture (NA) of the lens was 0.43 and the shortest
focal length were determined to be 0.43 and 18 mm, respectively.
To capture the light rays, fluorescence dye Rhodamin B (Signma-Aldrich, excita-
tion wavelength = 540 nm, emission wavelength = 625 nm) was diluted in the mix-
ture. Grayscale images were recorded with a CCD camera (Princeton Instruments,
Trenton NJ). To characterize the acoustic converging ability of the lens, the trans-
ducer was driven by a pulser/receiver (5072PR, Olympus) to generate acoustic emis-
sion. A hydrophone (HGL-0200, ONDA) with a circular sensing area of 0.2mm in
diameter was scanned laterally to map the acoustic pressure distribution from the
transducer. Figure 5.23b shows the measured optical rays and acoustic pressure dis-
tributions when the lens was flat and fully infused. We can see that the lens enabled
coaxial acoustic and optical confocal with a focal length of 19mm. As we expected,
the optical and pressure intensities were highly confined when the lens was fully
infused. However, the measured focal length was slightly longer than the theoretical
estimation since the alignment between the collimator/transducer and the lens was
not optimal.
The 1D profile along the dashed white line cross the acoustic focal point shown in
Figure 5.23b is given in Figure 5.24a, and was used to estimate the size of acoustic
250
Acoustic Intensity (a.u.)
205 µm
200
0.7 mm
150
Focal plane
100
50 130 µm
0
8 9 10 11 12 13 14
Distance (mm)
(a) (b)
40 µm 4.8 µm
300
205 µm
Photoacoustic Intensity (a.u.)
250 0 µm
130 µm
200
150
100
50 60 µm
0
100 150 200 250 300
Distance (µm)
(c)
FIGURE 5.24 (a) 1-D profile along the dashed white line crossing the acoustic focal point
shown in Figure 5.23(b). (b) B-scan images of the blade in the focal plane, and 130 μm and
205 μm off the focal plane. (c) 1D profile along the dashed line shown in (b).
focus. The FWHM was found to be 0.7 mm, which was slightly larger than the
theoretical estimation (0.6 mm) calculated by the following equation, due to some
misalignment of the transducer and the lens:
1.028 × f × c
Resolution ( = 6 dB ) = (5.12)
F×D
where f is the focal length, c is the sound velocity, F is the central frequency, and D
is element diameter of the transducer.
We employed the edge spread function (ESF) to estimate the lateral resolution
of this imaging system. The shape edge of blade (surgical scalpel, blade no. 10) was
imaged. One B-scan image contained 200 A-lines with an interval step of 2.5 μm,
and a total of 100 B-scan images were obtained by performing axial scanning with
a scanning step of 5 μm. Figure 5.24b shows the B-scan images when the blade was
in the focal plane, and 130 μm and 205 μm off the focal plane. From the profiles
given in Figure 5.24c, we found that the resolution in the focal plane is 4.8 μm, and
the resolutions in the out-of-focus planes are much poorer (40 μm in the plane and
130 μm off the focal plane, and 60 μm in the plane and 205 μm off the focal plane).
To evaluate the sensitivity with and without acoustic focused detection, a human
hair was embedded inside a tissue mimicking phantom with an absorption coefficient
of 0.007 mm–1 and a reduced scattering coefficient of 1 mm–1. In the first experi-
ment, the mixture of glycerol and water served as external liquid and enabled acoustic
and optical confocal. In the control experiment, pure water that enabled only opti-
cal converging was used. The MAP images (Figure 5.25a) and quantitative compari-
son (Figure 5.25b) show the significant improvement of sensitivity by using confocal
configuration.
In vivo experiments were carried out to demonstrate the potential biomedical
application of this technique. Rat was kept motionless using a breathing anesthe-
sia system and the body temperature was maintained at 37°C by a temperature-
controlling pad. The rat ear was gently depilated before tightly contacting with
the membrane sealed imaging window in the bottom of the tank. All procedures
were approved by the Institutional Animal Care and Use Committee (IACUC)
at the University of Florida. The total raster scanning points was 200 along each
transverse direction with an interval of 5 μm over a distance of 1 mm. The laser
pulse energy after the lens was measured to be 100 nJ. The experimental time for
a complete volumetric data collection was 40 minutes. After the data acquisition,
the rat fully recovered without any observable laser damage. The microvasculature
was clearly imaged as shown in Figure 5.26a where the top panel shows the MAP
image along axial direction, and the bottom panel gives a selected cross-section
image along the dashed white line in MAP. Figure 5.26b shows the 3D pseudo
color visualization of the vasculature.
We note that the application of this lens is not limited to ORPAM systems. If
we use an optical fiber bundle instead of single-mode optical fiber, we can build
an ARPAM imaging system with weakly confocal of the optical illumination and
Glycerol + water Pure water
Min Max
(a)
Glycerol + water Water

0.12
Photoacoustic Intensity (V)
0.1
0.08
0.06
0.04
0.02
0
0 20 40 60 80 100 120
Distance (um)
(b)
FIGURE 5.25 (a) MAP of human hair with and without acoustic focusing. (b) 1-D plots
along the dashed lines in Figure 5.25(a). Scale bar: 20 μm.
100 um
(a) (b)
FIGURE 5.26 (a) MAP image (top) and selected B-scan image (bottom) of the rat ear.
(b) Volumetric rendering of the vasculature of the imaged rat ear.
acoustic detection using a single liquid lens. In addition, the variable focal length of
the lens enables us to easily integrate fluorescence confocal microscopy with pho-
toacoustic microscopy. We also recognize that the central frequency of the current
ring-shaped transducer is not high, which led to relatively poor axial resolution and
low sensitivity to capillaries. In addition, the internal diameter of the transducer is
larger than the laser beam. A high-frequency ring-shaped transducer with a small
inner diameter can significantly improve the sensitivity and axial resolution of this
imaging system. Finally, a fast laser can significantly reduce the data acquisition
time.
6 Multimodal Approaches
When two or more than two different imaging modalities are physically combined
appropriately, the combined system can take full advantage of different modalities
to provide complementary information for most accurate clinical decision making.
This is evidenced by the combination of computed tomography (CT) and positron
emission tomography (PET) where CT and PET provide, respectively, structural and
physiological information of common tissue. In this chapter, we present examples of
combining photoacoustic tomography (PAT) with several other modalities includ-
ing diffuse optical tomography (DOT), fluorescence molecular tomography (FMT),
ultrasound imaging, and optical coherent tomography (OCT) to realize various PAT-
based multimodal approaches.
6.1 PAT/DOT
While PAT has been shown so far to be particularly powerful in imaging vascula-
tures and tissue-absorption related parameters such as hemoglobin and oxygenation,
PAT has limitations. For example, PAT can provide high-quality images only for
certain sizes of targets due to the limited detection band of an ultrasound transducer.
It is difficult to accurately detect a relatively large size target (e.g., ≥ 6 mm) using
a typical ≥1 MHz transducer. In addition, the limited directivity of a transducer
may lead to imbalanced image quality for the entire imaging area/volume. Yet, it
still remains a major challenge for PAT to recover tissue-scattering coefficients, an
important parameter for breast cancer detection (Jiang 2010).
DOT is capable of overcoming the above-mentioned limitations associated PAT,
although it has relatively low spatial resolution. Thus, a combination of PAT and DOT
appears to be an ideal approach for tissue imaging. Here we describe a combined
PAT/DOT system and demonstrate its ability using phantom and ex vivo experiments.
6.1.1 Material and Methods

6.1.1.1 System Description
Figure 6.1a presents the schematic of our PAT/DOT hybrid system. Detailed descrip-
tion of the DOT part has been reported elsewhere (Li et al. 2006). Briefly, light
generated from a diode laser is delivered sequentially by an optical switch and opti-
cal fiber bundles to 16 source positions. For each source position, diffused light is
detected by 16 detection fiber bundles coupled with detection units and data acquisi-
tion board and used for DOT image reconstruction.
For the PAT part, a pulsed light from a Nd:YAG or Ti:Sapphire laser with a 6-ns
pulse duration and 10-Hz repetition rate is sent to the object through a light deliv-
ery system. The laser-generated ultrasound signals are collected by 64 transducers,
169
(b)
Positive electrode #1
(a) PVDF Backing
Laser Fiber bundle
Optical swith
Positive electrode #2
16 fiber bundles Pre-amplifier 1–4

(sources)
Pre-amplifier 5–8 16 channel data
16 fiber bundles acquisition card
(detectors) Pre-amplifier 9–12
Detection units Pre-amplifier 13–16

Personal computer I
Data acquisition board

Light delivery system Pulsed laser
Personal computer II Trigger
FIGURE 6.1 (a) Schematic of the PAT/DOT system. (b) Photograph of the interface and
schematic of a single transducer.
which are connected via a mechanic switching system to a 16-channel preamplifier

and 16-channel data acquisition (DAQ) board. The sampling rate of the DAQ board
triggered by the laser is 50 MHz.
As shown in Figure 6.1b, the finished optical fibers/transducers/object interface
contains a homemade 64-element acoustic transducer array. We used commercial
PVDF film (110 µm) with silver electrode in both sides. The film was shaped into
32 rectangle units with a size of 5 mm × 30 mm. For each unit, a hole in the center
was drilled for the placement of a fiber bundle of DOT. The positive electrode was
divided into two parts with an equal size of 2.3 mm × 30 mm and the PVDF film
was attached to backing material with the same size as the film by transparent epoxy
resin. The positive and negative electrodes of a coaxial cable were connected to the
film using silver epoxy. A copper housing was used to shield electromagnetic noise
for each unit to improve the signal-to-noise ratio (SNR).
6.1.1.1.1 Light Delivery System

We initially tested a fiber-bundle-based light delivery approach and found it was
not feasible since the damage threshold for a typical fiber bundle is 20 mJ, which
is much lower than the at least 60 mJ needed for us to illuminate an object area of
3 cm 2 (calculated based on the maximum permissible exposure to light for human
tissue surface). Therefore, we developed a light delivery system by mounting three
prisms to a 2D step motor (Figure 6.2a). And a concave lens and ground glass
indicated by arrow 3 in Figure 6.2a were attached to the front of output end of
the light delivery system to extend the light beam to 3 cm 2. The output laser beam
Multimodal Approaches 171
Laser 12000
1 MHz
10000
Exam table
8000
1 6000
Prism 2 1.48 MHz
4000
Prism 1
2000
3 4 2 0.38 MHz
Prism 3 0
0 0.5 1 1.5 2 2.5 3 3.5 4
Frequency (MHz)
(a) (b)
Sensitivity
1
1
Directivity 0.9
0.9
0.8
0.8
0.7
Normalized Signal
Normalized Signal
0.7
0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 –30- 30-
0.2
0.1
0.1
0
0
–80 –60 –40 –20 0 20 40 60 80 0 10 20 30 40 50 60
Angle (Degree) Transducer Element
(c) (d)
FIGURE 6.2 (a) 3D schematic of the light delivery system. (b) Frequency response of a trans-
ducer. Dashed line shows the –6 dB level. (c) Directivity of a transducer. Dashed line shows the
–6 dB level. (d) Normalized sensitivity for all the elements in the transducer array.
was delivered from bottom of the exam table to the phantom. During an imaging
experiment, the light beam is scanned in 2D (arrows 1 and 2 in Figure 6.2a) via the
two step motors to realize the light illumination to cover a large area.
6.1.1.1.2 Frequency Response of Transducers

Ultrasound signal generated by photoacoustic effect on a sphere absorber is distributed
in bipolar shape, and the Fourier spectrum of this signal reveals the major frequency
components that the acoustic energy contains. Therefore, the frequency bandwidth of
the transducer determines the detectable range of target size. We used an impulsive wave
method described by Manohar et al. (2004) to measure the frequency response of each
element. Figure 6.2b shows that the –6 dB bandwidth of the transducer is from 380 kHz to
1.48 MHz and maximum frequency response is up to 2 MHz.
6.1.1.1.3 Directivity and Sensitivity of Transducers

Figure 6.2c presents the directivity of a transducer measured in the plane parallel
to the short side of the transducer. It was estimated to be ±30º based on the –6 dB
level. The methodology we used for the estimation was described by Ermilov et
al. (2009). The normalized sensitivity for the 64 elements is shown in Figure 6.2d,
where we see that the sensitivity lay in the range of 0.7 to 1.0. We used this set of
data to calibrate the measurements from each element for image reconstruction.
6.1.1.2 Quantitative Reconstruction Algorithms

The reconstruction methods for both PAT and DOT in our PAT/DOT system are finite
element (FE) based. For PAT, the FE-based reconstruction algorithms are described
detailed in Chapters 1 and 2. A fine mesh of 6285 nodes and a coarse mesh of 1604
nodes were applied in the PAT reconstruction, and the images were converged within
50 iterations using a parallel computer.
For DOT, both the absorption and scattering coefficient images were recovered
from the algorithms described in detail in Jiang (2010). In short, the algorithms use
a regularized Newton’s method to update an initial optical property distribution
iteratively in order to minimize an object function composed of a weighted sum of
the squared difference between computed and measured optical data at the medium
surface. The computed optical data (i.e., photon density) is obtained by solving the
photon diffusion equation with a finite element method. The core procedure in our
reconstruction algorithms is to iteratively solve the following regularized matrix
equation:
(ℑT ℑ+ λI ) q = ℑT (Φ( m ) − Φ(c ) ), (6.1)

where Φ is the photon density, I is the identity matrix, and λ can be a scalar or a
diagonal matrix. q = ( D1 , D2 ,..., DN , a ,1 , a ,2 ,..., a ,N )T is the update vec-
tor for the optical property profiles, where N is the total number of nodes in the finite
element mesh used and D is the diffusion coefficient. Φ( m ) = (Φ1(m ) , Φ(2m ) ,..., Φ(Mm ) )
and Φ(c ) = (Φ1(c ) , Φ(2c ) ,..., Φ(Mc) ), where Φ(i m ) and Φ(i c ), respectively, are measured
and calculated data for i = 1,2,...,M boundary locations. ℑ is the Jacobian matrix
that is formed by ∂Φ/∂D and ∂Φ/∂ a at the boundary measurement sites. In DOT,
the goal is to update the a and D or s distributions through the iterative solu-
tion of Equation (6.1) so that a weighted sum of the squared difference between
computed and measured data can be minimized. A single mesh of 700 nodes was
used, and the images were converged within 15 iterations in a 3 GHz PC with
1 GB memory.

Since detailed system performance for DOT has been evaluated elsewhere (Li
et al. 2006), here we focus on evaluating the PAT part of the PAT/DOT system on the
imaging depth, active imaging area, and multitarget imaging ability using the light
–10 –10
–8 –8
–6 3 –6
–4 –4 2
2
–2 –2
1
0 1 0
2 2
4 4
6 6
8 8
10 10
–10 –5 0 5 10 –10 –8 –6 –4 –2 0 2 4 6 8 10
(a) (b)
FIGURE 6.3 (a) PAT image of three separated metal wires (0.15 mm in diameter each)
embedded in the phantom background. (b) PAT image of two metal wires (0.15 mm in diam-
eter each) placed in the phantom background.
delivery system, and conduct a comparison between PAT and DOT in these areas.
We also test our PAT/DOT system using ex vivo tumor tissue embedded in a tissue-
mimicking phantom.
6.1.2.1 Spatial Resolution of the System

Spatial resolution of a PAT system is determined by several factors including the num-
ber of transducers, inter-element spacing, and sensitivity and frequency response of
the transducer as well as the reconstruction methods used (Andreev et al. 2003). Two
experiments were performed to estimate the best spatial resolution of our system. In
the first experiment, we put three thin metal wires (0.15 mm in diameter each) in the
center of a phantom background. The distance was 0.5 mm and 0.7 mm, respectively,
between wires 1 and 2 and between wires 2 and 3. In the second experiment, we placed
two metal wires into the phantom background, and the distance between the two wires
was 0.3 mm. The reconstructed PAT images from the two experiments are shown in
Figures 6.3a and 6.3b, respectively, where we note that the three targets can be clearly
distinguished (Figure 6.3a), while the two targets are not resolvable (Figure 6.3b).
Hence, we determine that the best spatial resolution of our system is 0.5 mm.
6.1.2.2 PAT Performance
Figure 6.4a shows the geometry of the tissue-mimicking phantom (Intralipid + India
ink) and the positions of five targets used for four phantom experiments. The param-
eters for the five targets are listed in Table 6.1. These experiments were performed
using the 532 nm Nd:YAG pulsed laser. The light beam was guided and extended
by our light delivery system to be an area source of 3 cm2 and the PAT images were
reconstructed by a delay and sum method. The light energy density at the surface of
the phantom was 20 mJ/cm2.
3
4
1 5
2 Phantom
(a)
–20 –20
–15 –15
–10 –10
–5 –5
1
1 0
0
5 5
10 10
15 15
20 20
–20 –15 –10 –5 0 5 10 15 20 –20 –15 –10 –5 0 5 10 15 20
(b) (c)
–40 –40
–30 –30
–20 –20 4
3
–10 –10
0 0
10 10 5
20 20
30 30
2
40 40
–40 –30 –20 –10 0 10 20 30 40 –40 –30 –20 –10 0 10 20 30 40
(d) (e)
FIGURE 6.4 (a) Geometry and positions of targets used for four phantom experiments.
(b) PAT image of target 1 without any depth. (c) PAT image of target 1 positioned at 22 mm
beneath the surface. (d) PAT image of target 3 (30 mm off-center positioned). (e) Fused image
of multiple targets.
TABLE 6.1
Exact Size, Location (Off Center), Depth, and Absorption and Reduced
Scattering Coefficients of the Five Targets
Size (mm) Location (mm) Depth* (mm) μa (mm–1) μ′s (mm–1)
Target 1 2 0 0 (22) 0.028 2
Target 2 2 20 5 (15) 0.028 2
Target 3 3 30 5 (10) 0.028 2
Target 4 3 20 5 (15) 0.028 2
Target 5 1 20 5 (15) 0.028 2
* The maximum experimental imaging depth for certain cases is given in the parentheses.
Figures 6.4b and 6.4c show the PAT images of target 1 located at 0 and 22 mm
below the phantom surface, respectively. While the target is clearly better imaged
with 0 mm depth, the target with 22 mm depth is still detectable. The active imag-
ing area for PAT is validated by the result shown in Figure 6.4d where target 3
(30 mm off-center positioned) is detected. Due to the limited directivity of trans-
ducers, we note that the shape of reconstructed target was distorted and stronger
artifacts are seen around the target. We also tested and found that when a target
was placed at a >30 mm off-center position, the image quality became unaccept-
able. For the single target experiments, the light beam was directly delivered to
the target area without scanning. The image shown in Figure 6.4e demonstrated
the ability of imaging multiple targets using the developed light delivery system.
Three targets (2, 4, and 5) separated with a large distance were imaged by scan-
ning the laser beam nine (3 × 3) times to cover the whole phantom surface. The
nine images obtained from the nine laser beam positions were then fused into a
single image shown in Figure 6.4e.
6.1.2.3 PAT and DOT Comparison

Figure 6.5 presents the reconstructed quantitative absorption coefficient images by
PAT and absorption and scattering coefficient images by DOT for some of targets 1
to 5 and for a relatively large target, target 6 positioned 20 mm off-center (radius =
5 mm, absorption coefficient = 0.028 mm–1, and scattering coefficient = 4.0 mm-1).
These quantitative images were recovered using our finite element based PAT and
DOT. Figure 6.5 also gives the reconstructed optical property profiles depicted along
one cut line crossing through the center of the target (dashed line) for each case in
comparison with the exact property values (solid line).
From Figure 6.5a, we see that target 5 (radius = 1 mm) is accurately recovered by
PAT in terms of its size, position, and absorption coefficient value, while it is not detect-
able by DOT in both the absorption and scattering images. For target 2 (radius = 2 mm),
again PAT can accurately reconstruct its size, position, and absorption coefficient value
40 PAT PAT y = 2 mm
Exact
0.02 0.025 Recovered
20
0.02
0.015
µa 0.015
0
0.01
0.01
–20 0.005
0.005
0
–40 8 10 12 14 16 18 20 22 24 26
–40 –20 0 20 40 x (mm)
40 DOT DOT
y = 2 mm
0.018 Exact
Recovered
20 0.016 0.025
0.014
0.02
0 0.012 µa
0.015
0.01
–20 0.008 0.01
0.006
–40 8 10 12 14 16 18 20 22 24 26
–40 –20 0 20 40
x (mm)
40 DOT DOT y = 2 mm Exact

1 2
Recovered
20 1.8
0.9
1.6
0 0.8 µs' 1.4
1.2
–20 0.7
1
0.8
–40 0.6
–40 –20 0 20 40
8 10 12 14 16 18 20 22 24 26
x (mm)
(a)
FIGURE 6.5 Quantitative PAT and DOT images of target 5 (a), target 2 (b), target 3 (c), and
target 6 (d). (continued)
(Figure 6.5b). In this case, DOT can detect the target from both the absorption and
scattering images; however, the values of both absorption and scattering coefficients
are significantly underestimated. Figure 6.5c shows that target 3 (radius = 3 mm) is
quantitatively recovered by both PAT and DOT while we note an overestimated target
size by DOT.
Interestingly, the largest target 6 (radius = 5 mm) is poorly recovered by PAT
(Figure 6.5d), due to the loss of low-frequency signals given the limited frequency
response of the transducers. Here we see that only the edge of the target is recovered
by PAT and the reconstructed absorption coefficient value is considerably underesti-
mated. DOT in this case provides accurate recovery of the target in terms of its size,
position, and absorption and scattering coefficient values.
x = 3 mm
40 PAT PAT
Exact
30 0.02 0.025
Recovered
20
0.02
10 0.015
0 µa 0.015
0.01
–10 0.01
–20
0.005 0.005
–30
–40 0
–40 –20 0 20 40 –28 –26 –24 –22 –20 –18 –16 –14 –12 –10
y (mm)
40 DOT DOT x = 4.2 mm

0.015
30 Exact
0.025
0.014 Recovered
20
0.013 0.02
10
0.012
0 0.015
0.011 µa
–10
0.01 0.01
–20
0.009 0.005
–30
0.008
–40 0
–40 –20 0 20 40 –26 –24 –22 –20 –18 –16 –14 –12 –10 –8
y (mm)
40 DOT DOT x = 4.2 mm
2 Exact
30 1.4
Recovered
20 1.3 1.8
10 1.2 1.6
0 1.1 µs' 1.4

–10 1 1.2
–20 0.9
1
–30 0.8
0.8
–40
–40 –20 0 20 40 –28 –26 –24 –22 –20 –18 –16 –14 –12 –10 –8
y (mm)
(b)
FIGURE 6.5 (continued) Quantitative PAT and DOT images of target 5 (a), target 2 (b),
target 3 (c), and target 6 (d). (continued)
Overall from the results shown in Figure 6.5 and the summary in Table 6.2, PAT
can provide both qualitatively and quantitatively better images than DOT when the
target size is equal to or small than 3 mm in radius, while DOT can offer better
image quality when the target size is larger than 3 mm in radius. We also notice that
the smallest detectable target size is 4 mm for DOT given the experimental condi-
tions used in this study.
6.1.2.4 Ex Vivo Experiment

For the ex vivo experiment, a tumor was removed from a rat bearing a 4T1 tumor
of approximately 3.5 mm in radius and embedded in the phantom background.
In this case, 730 nm pulsed light from a Ti:Sapphire laser was employed and
4
× 10
40 PAT PAT x = –22 mm
16 Exact
30
14 0.025 Recovered
20
12 0.02
10
10
0 µa 0.015
8
–10
6 0.01
–20 4
–30 0.005
2
–40 0
–40 –20 0 20 40 4 6 8 10 12 14 16 18 20 22
y (mm)
40 DOT DOT x = –24 mm

30 0.025 Exact
0.025 Recovered
20
10 0.02
0.02
0
µa
–10 0.015
0.015
–20
–30 0.01 0.01
–40
–40 –20 0 20 40
0.005
4 6 8 10 12 14 16 18 20
40 DOT
DOT x = –24 mm
30 2 Exact
1.7
Recovered
20 1.8
1.6
10
1.5 1.6
0 µs'
1.4
–10 1.4
1.3
–20 1.2
–30 1.2
1.1 1
–40
–40 –20 0 20 40 4 6 8 10 12 14 16 18 20
y (mm)
(c)
target 3 (c), and target 6 (d). (continued)
provided an energy density of 20 mJ/cm 2 at the phantom surface. Figure 6.6 shows

the recovered absorption and/or scattering images, where we see that the tumor
is detected by both PAT and DOT. The absorption coefficient of the tumor recov-
ered by PAT is consistent with that by DOT. Through quantitative analysis of ex
vivo PAT and DOT images, the size of tumor (3 mm in radius) estimated by the
absorption image of PAT and by the scattering image of DOT is consistent with
the actual tumor size, while it is overestimated by the absorption image of DOT
to be 5 mm in radius.
–3
×10 0.03 y = 17 mm
40 PAT PAT Exact
12 Recovered
0.025
20 10
0.02
8
µa 0.015
0
6
0.01
4
–20
0.005
2
–40 0
2 4 6 8 10 12 14 16 18 20 22
–40 –20 0 20 40
x (mm)
40 DOT DOT y = 17 mm
Exact
0.025
0.025 Recovered
20
0.02
0.02
0 µa
0.015 0.015
–20
0.01 0.01
–40
–40 –20 0 20 40 2 4 6 8 10 12 14 16 18 20 22
x (mm)
DOT DOT y = 17 mm
40 4 4 Exact
3.5 3.5 Recovered
20
3 3
µs'
2.5 2.5
0
2
2
–20 1.5
1.5
1
–40 1
2 4 6 8 10 12 14 16 18 20 22
–40 –20 0 20 40
x (mm)
(d)
target 3 (c), and target 6 (d).
6.2 PAT/FMT
Fluorescence molecular tomography (FMT) has received particular attention in
recent years (Jiang 2010), due to its remarkably high sensitivity as well as the increas-
ing availability of fluorescent dyes and probes, which provide a variety of molecular
signatures. However, the major limitation of FMT is its low spatial resolution, espe-
cially in reflection mode. While PAT has high spatial resolution, it has relatively lower
sensitivity compared with FMT when near-infrared (NIR) fluorescent dyes such as
indocynine green (ICG) are used (see more discussion about molecular or contrast
TABLE 6.2
Exact and Reconstructed Target Size and Absorption and
Reduced Scattering Coefficients for the Phantom Experiments
Size (mm) μa(mm–1) μ′s (mm–1)
Case 1 Exact 2 0.028 2
PAT 2 0.026 NA
DOT NA NA NA
PAT 4 0.025 NA
DOT 10 0.017 1.1
PAT 6 0.024 NA
DOT 10 0.024 1.8
Case 4 Exact 10 0.028 4
PAT 8 0.015 NA
DOT 12 0.028 4
agents based PAT in Chapter 7). While many contrast-agents-based PAT methods
use nanoparticles (Chapter 7), some of them use fluorescent dyes as dyes are usually
strong absorber. In this case, a dual-imaging modality of PAT and FMT can be real-
ized where PAT offers excellent spatial resolution and FMT provides high sensitivity,
making the combination a better molecular imaging approach. Here we describe a
combined 3D PAT and FMT in reflection-mode where the common contrast agent is
ICG dye.
6.2.1 Methods
The PAT system is schematically shown in Figure 6.7a. A tilted fiber bundle deliv-
ers laser light from a Ti:Sapphire tunable laser providing a generally homogenous
illumination. The incident intensity is up to 20 mJ/cm2 on the object surface, which is
100% of the American National Standards Institute limit. The laser source has a rep-
etition rate of 10 Hz, and the pulse duration is about 6ns. Different laser wavelengths
were chosen for the different ICG concentrations in the phantom experiments so
that the absorption peak of ICG could be reached. A 4 MHz immersion ultrasound
transducer was scanned by a linear moving stage along x and y directions with a step
size of 1.27 mm and 1.00 mm, respectively. The generated photoacoustic waves were
averaged 50 times for each position, taking about 40 mins for a complete area scan.
The complex signal was amplified by a commercial preamplifier and acquired by a
high-speed PCI data acquisition board. A LabView program controlled the entire
system for data acquisition. The 3D images were reconstructed using the delay-and-
sum algorithm described in Chapter 1.
x = 1.2 mm
× 10–3
16
40 PAT PAT Recovered
14
0.015
12
20
10
0.01 µa 8
0
6
0.005 4
–20
2
0
–40 0
–40 –20 0 20 40 10 15 20 25
y (mm)
(a)
× 10–3 × 10–3 x = 1.2 mm

40 DOT 14
DOT Recovered
12
12
20
10 10
0 µa
8 8
–20 6
6
–40 4 4
–40 –20 0 20 40 5 10 15 20 25
y (mm)
(b)
x = 1.2 mm
40 DOT 1.4
1.4 DOT Recovered
20 1.3 1.3
1.2 1.2
0
µs'
1.1 1.1
–20
1
1
–40 0.9
–40 –20 0 20 40 5 10 15 20 25
y (mm)
(c)
FIGURE 6.6 Reconstructed ex vivo absorption image by PAT (a), and absorption (b) and
scattering (c) images by DOT.
Diode laser y
Pulsed laser Transducer x
CCD
y
x
Filter
Acoustic waves
Multiply scattered
Target Target fluorescence light
(a) (b)
FIGURE 6.7 Schematic of the PAT (a) and FMT (b) system.
The FMT setup is shown in Figure 6.7b. A continuous-wave (CW) 785-nm diode

laser was mounted on a linear moving stage to deliver the excitation light through a
convex lens to preselected source positions at an incident power of 20 mW. A 1024 ×
1024 pixel CCD camera was placed in front of the phantom. The angle between the
laser beam and the camera was about 15°. An 830 nm band-pass filter was employed
in front of the camera to eliminate the excitation light. For each experiment, the
exposure time for the CCD camera varied and pixel binning was used for improved
SNR. In this study, a total of 36 source positions and 45 detector positions were used
covering an area of 20 × 20 mm in the x–y plane.
The fluorescence images obtained were reconstructed using an iterative finite ele-
ment based algorithm that is similar to the DOT algorithm described in Section
6.1.1.2 (Equation (6.1)) except that the photon density for excitation is replaced by
that for fluorescent light.

ICG absorbs light primarily in the range of 600 and 900 nm and emits fluores-
cence light from 750 to 950 nm. It has a molar extinction coefficient of 2 × 104 ~
1 × 105 cm–1/M, depending on the excitation wavelength (Landsman et al. 1976),
and a quantum yield of ~0.027 (Philip et al. 1996), when the concentration is less
than 1 mM in water. To comparatively evaluate the performance of FMT and
PAT, the same ICG-containing phantoms were used for both modalities, and the
reconstructed FMT and PAT images were compared in terms of the image pattern/
morphology, spatial resolution, and sensitivity. The background phantom was com-
posed of Intralipid, India ink, and agar powder providing μa = 0.01/mm and μ′s = 0.7/
mm. The targets were different in size, number, and ICG concentration for different
experiments. The reconstructed volume was 2 cm × 2 cm × 2 cm for FMT, while for
PAT the reconstructed volume was 2 cm × 2 cm × 1.7 cm located 3 mm below the
surface. When comparing the image pattern and spatial resolution, a wavelength of
810 nm and a laser energy density of 3 mJ/cm2 on the phantom surface were used
for PAT. For sensitivity study, different wavelengths at a laser energy density of
20 mJ/cm2 were used for PAT, while one excitation wavelength at a fixed incident
laser power was used for FMT.
6.2.2.1 Comparison of Image Pattern

One cylindrical solid target with a diameter of 2 mm and a height of 1 mm was
embedded in the center of a cubic background phantom. ICG concentration of the
target was 100 µM. Figure 6.8 shows the PAT and FMT images at selected z-x (y =
10 mm) and x–y (z = 7 mm) slices with the exact location of the targets indicated by
the white rectangle/circle for comparison. We note that the target was clearly detected
by both methods. However, the two methods show different patterns. For example,
from Figure 6.8a, the PAT recovered target is clearly identified with high resolution.
It is also noted that strong signal is observed at the top and bottom surfaces of the
target and that the other boundaries and inner volume were not reconstructed well.
In contrast, the FMT reconstructed target size is significantly overestimated with the
center position shifted by 1 mm. Although the spatial resolution is not high compared
with that in the PAT images, the FMT images show clear target reconstruction. The
incomplete PAT recovery of target along z-direction (Figure 6.8a) can be explained
by the insufficient number of detector positions given the limited angle reflection-mode
used, as well as the lack of low-frequency information from the transducer used.
0 4000
0
4
5 3000
8
X
PAT
2000
Z 10
12
15 16 1000
20
0 5 10 15 20 0 4 8 12 16 20
X Y
(a) (b)
0 0
0.3
4 4 0.25
8 8 0.2
FMT
Z X 0.15
12 12
0.1
16 16
0.05
20 20 0
0 4 8 12 16 20 0 4 8 12 16 20
X Y
(c) (d)
FIGURE 6.8 PAT and FMT images for a 100 µM ICG target at 7 mm depth. (a) and (b) z-x
image, x–y PAT image; (c) and (d) corresponding FMT image. The white rectangle/circle
indicates the exact target location.
6.2.2.2 Comparison of Spatial Resolution

Three sets of phantom experiments were conducted. Two cylindrical targets (1 mm
in diameter and 1 mm in height) were located at a 7 mm depth. The separation
between the centers of the two targets differed by 2 mm, 4 mm, and 6 mm, respec-
tively, in the horizontal plane. Each target contained 100 µM ICG. Figure 6.9 shows
the reconstructed PAT and FMT images for two targets with different horizontal
2 mm 3.5 mm
0 0
350
500
5 5 300
400
10 10 250
300
15 15 200
200
20 20 150
0 10 20 0 10 20
(a) (b)
7 mm
0 350 0 0.2
5 300 5
0.15
10 250 10
0.1
15 200 15
20 150 20 0.05
0 10 20 0 10 20
(c) (d)
0 0.2 0 0.2
5 5
0.15 0.15
10 10
0.1 0.1
15 15
20 0.05 20 0.05
0 10 20 0 10 20
(e) (f )
FIGURE 6.9 Reconstructed images with different target separations. (a)–(c) PAT image;
(d)–(f) FMT image.
separations where the exact target positions are indicated by the black circles. We
see that the two targets at the three separations are clearly reconstructed by both
methods. However, it is noted that the two targets are not fully separated by FMT
when they are only 2 mm apart (Figure 6.9d), but they are clearly separated by PAT
(Figure 6.9a).
6.2.2.3 Comparison of Sensitivity
Nine sets of phantom experiments were conducted to evaluate the sensitivity using
three different ICG concentrations (100 µM, 10 µM, and 2 µM) and three different
target depths. To achieve the maximum absorption at different ICG concentration,
three different wavelengths (698 nm, 780 nm, and 790 nm) were used for PAT. For each
ICG concentration, the targets (2 mm in diameter and 1 mm in height each) were buried
at 7 mm, 12 mm, and 17 mm below the surface, respectively. The targets were 2 mm in
diameter and 1 mm in height. Figure 6.10 shows the reconstructed PAT images along
x-z plane (y = 10 mm), while Figure 6.11 gives the corresponding FMT images. We
immediately note that the PAT images provide much more reliable target position and
size because of the high spatial resolution compared to its FMT counterpart. From
Figure 6.10, it can be seen that as the ICG concentration decreases or the target posi-
tion goes deeper, the reconstructed PAT image has greater artifacts and decreased
image quality. While similar trend is seen from the FMT images, we note that FMT
has clearly greater sensitivity compared to PAT for detection of ICG fluorochromes.
For example, the target with 10 µM ICG at 17 mm depth or 2 µM ICG at 12 mm depth
100 µm 10 µm 2 µm
5 x 1500 5 80 5 50
10 60 40
10 z 1000 10
7 mm
40 30
15 500 15 15 20
20 10
20 20 20
0
0 10 20 0 10 20 0 10 20
5 5 40 5 40
200 10 30
30
12 mm
10 10
15 20 20
15 100 15
10 10
20 20 20
0 0 0
0 10 20 0 10 20 0 10 20
5 140 5 5
120 60
40
17 mm
10 100 10 10
80 40
15 60 15 15 20
40 20
20
20 20 20
0
0 10 20 0 10 20 0 10 20
FIGURE 6.10 PAT images for a target with different ICG concentrations at different depths.
100 µm 10 µm 2 µm
0
4
8
7 mm
Z
12
16
20
0 4 8 12 16 20
X
12 mm
17 mm
0.1 0.2 0.3 0 0.1 0.2 0 0.2
FIGURE 6.11 FMT images for a target with different ICG concentrations at different
depths. The black block indicates the true target position.
is imaged by FMT (Figure 6.11), while the corresponding target is not detectable by

PAT (Figure 6.10). Such difference in sensitivity is due to the fact that both the target
and background generated ultrasound upon the optical absorption in the case of PAT,
while only the target produces NIR fluorescent emission in the case of FMT.
6.3 PAT/ULTRASOUND
While PAT can provide high-resolution tissue absorption coefficient images and sub-
sequently derived functional parameters such as hemoglobin and oxygenation, it often
lacks certain tissue structural information. One way to compensate this is to combine
ultrasound (US) imaging with PAT. This is actually the most convenient way to realize
a combined PAT and US since both share the common detection channel for ultra-
sound signal collection. We illustrate such a PAT/US combination using intravascular
Ti: Sapphire laser

Trigger
Data acquisition
board
Angle polished MM fiber end
Pulser & receiver
1 mm
Micro transducer
Lens PA or US signal
Multi-mode (MM) fiber

Mirror
Blood vessel
FIGURE 6.12 Schematic of a combined PAT/ultrasound system for intravascular imaging.
(IV) imaging system as an example (see Figure 6.12). In this PAT/US combined system,
light beam from a pulsed Ti:Sapphire (750–1700 nm) laser is reflected by a mirror
and focused into a multimode (MM) optic fiber via a lens. The light is delivered into a
1.5 mm-diameter miniature imaging probe consisting of the MM fiber with angle pol-
ished end and a 20 MHz central frequency microtransducer with a size of 0.7 × 0.5 mm2 (see
the insert photograph in Figure 6.12). The angle polished fiber end ensures the delivery
of light to the wall of blood vessel, and the light-induced ultrasound is collected by the
microtransducer. The photoacoustic (PA) signal is received by the receiver part of the
pulser/receiver and digitized by the data acquisition board/computer. An ultrasound
pulse is sequentially generated by the pulser and delivered into the miniature imaging
probe. In this case, the microtransducer acts as both the excitation source and detector,
and the received ultrasound signal goes through the pulser/receiver and is digitized
by the data acquisition board/computer. A 360º rotation of the imaging probe or the
sample allows a full scan of the blood vessel wall.
Figure 6.13 shows the intravascular PA (IVPA) and ultrasound (IVUS) images
obtained from two ex vivo human blood vessels with plaques (see the photographs;
the plaques are indicated by white arrows). While both IVPA and IVUS can iden-
tify the plaques in both cases, the images generated from the two different contrast
mechanisms certainly complement each other, which can be seen more striking from
the PAT/US fused images. We will have more discussion on IVPA in Chapter 8
where the unique ability of IVPA for quantitatively imaging biochemical composi-
tions of the plaques is demonstrated.
IVPA IVUS Fused
IVPA IVUS Fused
FIGURE 6.13 Intravascular photoacoustic (IVPA) and ultrasound (IVUS) images obtained
from two plaque-containing human blood vessels. Arrow indicates the location of plaque.
6.4 OPTICAL-RESOLUTION PAM/OCT
As a final example in this chapter, we present the combination of ORPAM and OCT
based on a miniature probe of only 2.3 mm in diameter, which will be ideal for endo-
scopic and intravascular imaging. The optical paths of both the ORPAM and OCT
are built upon a single-mode fiber, a miniature GRIN lens and two microprisms,
enabling the two modalities to scan the same tissue area. The self-focusing ability of
the GRIN lens results in a highly focused light beam for OCT and ORPAM, yielding
a high lateral resolution of 15 μm for both modalities.
6.4.1 Materials and Methods

Figure 6.14a shows the schematic of the integrated probe. In this probe, the illu-
mination beams for both OCT and ORPAM are coupled into one tip of a sin-
gle-mode fiber (SMF-28e+, Thorlabs) with 0.14 NA and 0.9 mm outer diameter.
The other tip of the fiber is cut with an 8° angle to minimize back-reflection.
Optical UV glue is used to connect the fiber tip with the GRIN lens resulting in
a 5 mm working distance. Two microprisms (one without coating and the other
one coated with aluminum) are glued together, and a miniature unfocused ultra-
sound transducer with 10 MHz central frequency and 2 mm aperture is mounted
on the top of the cubic prism pair. The light beams are focused by the GRIN
lens and reflected by the thin aluminum film to the tissue surface. The generated
ultrasound waves transmit through the cubic prism group and are detected by
the transducer. The back-scattering photons from the tissue are reflected by the
aluminum film and coupled into the same fiber through GRIN lens. Figure 6.14b
shows a photograph of the probe, where both the cubic prism pair and GRIN lens
Transducer Bottom view

Stainless steel tube
Single-mode fiber
Grin lens Prism group
1 mm
Tissue
(a) (b)
FIGURE 6.14 Schematic (a) and photograph (b) of the hybrid probe.
have 0.7 mm in diameter. A stainless steel tube with 1.0 mm in diameter is used
to protect the light path (i.e., the single mode fiber, GRIN lens and cubic prism
pair). The probe is glued to the transducer and protected by another bigger stain-
less steel tube (2.3 mm in diameter).
The probe is mounted on a 3D linear stage (Figure 6.15) and scanned in 2D for 3D
image formation. The light beam for ORPAM is generated from a pulsed Nd:YAG
laser with 532 nm wavelength and a 10 Hz repetition rate. Two neutral density filters
are used to attenuate the light energy, and a small iris is utilized to provide homoge-
neous light distribution. The shaped light beam is focused by a convex lens, which
passes through a 50 μm pinhole for spatial filtering and is then coupled into a 2 × 1
beam coupler. The detected acoustic waves are amplified by two wideband ampli-
fiers and digitized by a data acquisition board (NI5152, National Instrument) at a
sampling rate of 250 MS/s.
Iris
ND L1 L2
L3
Nd: YAG pulsed laser
Pinhole
SLD source
Circulator 2×1
2 × 2 BP
Probe
PD RSOD
2D linear stage
PC
Pre-amplifier
FIGURE 6.15 Schematic of the integrated ORPAM and OCT system. ND: neutral density,
L1: lens 1, L2: lens 2, L3: lens 3, BP: beam splitter, PD: photodiode.
70 100
OCT ORPAM
60
80
OCT Profile (a.u.)
50
PA Signal (a.u.)
40 60
30
40
20
10 20
0
0
0.1 0.15 0.2 0.25 0.3 0.35 0.1 0.18 0.27 0.35
x-axis (mm) x-axis (mm)
(a) (b)
FIGURE 6.16 Resolution test for ORPAM and OCT. 1-D profile of OCT (a) and ORPAM
(b) of Group 6, Element 1.
A broadband light source (DenseLight, DL-BX9-CS3159A) with a center wave-

length of 1310 nm is employed for OCT. The light source has a FWHM of 75 nm,
providing an axial resolution of 10 μm in air. The broadband light is split into the
reference arm and the sample arm by a beam splitter and coupled into the same 2 × 1
beam coupler used for ORPAM. The depth scanning from 0 to 1.6 mm at the refer-
ence arm is realized by a rapid scanning optical delay line (RSOD) coupled with a
galvanometer scanning at 1 kHz. The OCT signal is then detected by a balanced
photodetector, whose output is acquired and stored by a DAQ card. The sensitivity
of the system is measured to be 74 dB.
The lateral resolution is determined by imaging a selected part of an USAF 1951
resolution test target. Figures 6.16a and 6.16b show 1-D profile of three bars where
the smallest resolvable bar spacing is 15 µm (group 6, element 1). The three bars
can be clearly identified by both OCT and ORPAM, indicating a lateral resolution
of better than 15 µm. In tissue imaging, the scattering will reduce the spatial reso-
lution of this probe; however, when the sample is optically thin, the degradation of
lateral resolution is not significant (Xi et al. 2013). The imaging depth of ORPAM
is estimated to be 1.5 mm by imaging a pencil lead embedded in a turbid tissue
mimicking phantom.
In our current probe, the size of cubic prism pair (0.7 mm in diameter) is smaller
than the aperture of the transducer (2 mm in diameter). Hence, the active area of
transducer is divided into two areas: one area indicated by A1 in Figure 6.17a receives
the photoacoustic signal transmitting through the cubic prism pair; the other area
(A2) collects the photoacoustic signal transmitting through water without going
through the prism pair. Due to the different sound velocities in glass (3962 m/s) and
water (1480 m/s), the photoacoustic signal received by A1 will arrive earlier than
that received by A2 as indicated in Figure 6.17b. To compensate this difference, we
calculated the time shift between the signals received by A1 and A2, and used the cal-
culated time shift to calibrate the final signals for imaging. As shown in Figure 6.17c,
0.4
Without calibration
PA Intensity (Volt)
0.2 PA1 PA2
0
A2 0 1 2 3
z-axis (mm)
A1 (b)
0.4
With calibration
PA Intensity (Volt)
(a)
0.2
0
0 1 2 3
z-axis (mm)
(c)
FIGURE 6.17 Signal calibration. (a) Schematic of the probe (bottom view). (b) and (c):
Typical A-line (Hilbert Transform of raw data) without (b) and with (c) calibration.
the calibration allowed us to add the two signals together, resulting in improved SNR
for image formation.

To demonstrate the microscopic imaging ability of this dual-mode probe, we chose
to image the ear of a mouse. Before starting the experiments, the hair on the ear
was gently removed using a human hair-removing lotion. The mouse was placed
on a homemade animal holder and was anesthetized with a mixture of ketamine
(85 mg/kg) and xylazine. After the experiments, the mice were sacrificed using
the University of Florida Institutional Animal Care and Use Committee (IACUC)-
approved techniques. Strict animal care procedures approved by the University of
Florida IACUC and based on the guidelines from the NIH for the Care and Use of
Laboratory Animals were followed. The laser exposure was 25 mJ/cm2 at the optical
focus point which is higher than the ANSI laser safety limit (20 mJ/cm2), but still
below the reported skin damage threshold (Zharov et al. 2006). The scanning step
size was 6 µm along the x–y plane.
The top rows of Figures 6.18a and 6.18b show the maximum amplitude projec-
tion (MAP) images of ORPAM and OCT. OCT and ORPAM visualize different
tissue structures—ORPAM clearly maps the microvasculature, while OCT images
100 µm
ED
BV
CT D
FIGURE 6.18 In vivo imaging of mouse ear by the integrated probe. MAP image (top) and
cross-section (bottom) of ORPAM (a) and OCT (b). ED: epidermis, CT: cartilage, D: dermis,
BV: blood vessel.
the sebaceous gland with high resolution. The bottom rows (cross-section of tissue)
of Figures 6.18a and 6.18b show the benefits of combining these two modalities more
clearly. The ear’s thickness in the OCT image is from 400 to 600 µm. The dermal
structure and the sebaceous gland are clearly observed. In the cross-sectional OCT,
we can identify epidermis, dermis, and cartilage as indicated with arrows. We have
observed that ORPAM is good at locating micro vessels in the ear with limited sur-
rounding tissue information.
The SNR of the ORPAM is 25dB, which is lower than that of conventional
ORPAM. There are several reasons contributing to the reduced SNR: (1) A 10 MHz
ultrasound transducer was used in this probe, while it is known that the strongest
generated acoustic signal lies between 30 to 70 MHz; (2) the transducer is flat, result-
ing in reduced sensitivity compared with a focused transducer commonly used in
conventional ORPAM; and (3) during the experiments, we used an 8 bit resolution
DAQ card, which can resolve only signals that are larger than 40 mV. As a result, we
lost some signals from small capillaries. These limitations, however, can be easily
overcome using a miniature focused high-frequency transducer (>30 MHz), which
will in turn improve the axial resolution of ORPAM and make the whole probe
smaller as well. Using a high-resolution DAQ card will also help solve the afore-
mentioned problems as well as (4) the actual resolution of a GRIN probe tends to
be worse than this theoretical limit, because the effective NA of GRIN lenses is
degraded owing to the spatial aberrations in the lenses and misalignment between
optical components. For future improvements, we also need a suitable micromotor
to implement an internal scanning mechanism and a faster laser to reduce the scan-
ning time. Finally, we can replace the current slow time-domain (TD) OCT with fast
frequency-domain (FD) OCT. These are necessary steps towards the clinical evalu-
ation of the ORPAM/OCT.
7 Contrast Agents–Based
Molecular Photoacoustic
Tomography
In photoacoustic tomography (PAT), the interaction of visible and near-infrared

(NIR) light with tissue is dominated by absorbing chromophores including HbR,
HbO2, and H2O and scattering particles including the cell nuclei and membranes.
The sensitivity and specificity of PAT to visualize a pathological disorder are gov-
erned by its contrast: the ability of the disease to differentially absorb or scatter light
compared with the surrounding healthy tissue. This native or endogenous contrast
may not be sufficient, and in most cases, the interactions of light with tissue are not
disease specific. Therefore, there is a role for exogenously administered contrast-
enhancing agents that have affinity for the disease site through biochemical interac-
tions, providing not only sensitive but also disease-specific signals. The application
of contrast agents is able to increase the sensitivity and specificity of PAT and render
it a powerful molecular imaging tool. Both organic dyes such as indocyanine green
(ICG) (see Section 6.2) and inorganic nanoparticles are good candidates as photo-
acoustic contrast agents. In this chapter, we focus our discussion on several types of
nanoparticles as PAT contrast agents.
7.1 GOLD NANOPARTICLES
Gold nanoparticles have become a prime candidate for PAT due to their unusual
optical properties and inherent biocompatibility. The intense scattering and absorp-
tion of light that occurs under the plasmon-resonant condition coupled with the abil-
ity to tune the resonance into the NIR by manipulating the aspect ratio make gold
nanoparticles attractive as contrast agents for optical imaging techniques. Further,
gold-protein chemistry is well developed and several bioconjugation protocols are
available in the literature (Grobmyer and Moudgil 2010), which allows the combina-
tion of the targeting functionality of antibodies with such gold nanoparticles. The
inertness and biocompatibility of gold in general hold promise for the use of gold
nanoparticles for in vivo imaging applications.
Figure 7.1 shows the TEM image and optical absorption spectra for 20 nm and/or
50 nm gold nanoparticles. We see that these gold nanoparticles exhibit strong plasma
resonance peak around 530 nm. The molar extinction coefficient of the nanoparticles
was determined to be 1 × 109 M–1cm–1, about three orders of magnitude larger than
that of a typical organic molecule. Here phantom experiments were conducted to
demonstrate gold nanoparticles enhanced PAT using the imaging system described
193
0.03
20 nm
Absorption Coefficient (1/mm)

0.025 100 nm
0.02
0.015
0.01
0.005
0
400 500 600 700 800
Wavelength (nm)
(a) (b)
FIGURE 7.1 (a) TEM image of gold nanoparticles having an average size of 20nm. (b) Optical
absorption of 20nm (solid line) and 50nm (dashed line) nanoparticles.
in Chapter 1. The nanoparticles concentration used in our experiments ranged from

0.125 to 0.5 nM, which gave an absorption coefficient of the object from 0.0125 to 0.05
mm–1. In these experiments, we first embedded one or two nanoparticle-containing
objects (3mm in diameter) in a 10 or 25 mm-diameter solid cylindrical phantom (1%
Intralipid + India ink + distill water + Agar powder). We then immersed the object-
bearing solid phantom in a 110 mm-diameter water background. The images were
reconstructed using the FE-based algorithms described in Chapters 1 and 2.
Figures 7.2a to 7.2d show the reconstructed photoacoustic images from four sets
of experimental data. Figure 7.2a gives the image of one object containing 0.25 nM
nanoparticles, while Figures 7.2b, 7.2c, and 7.2d present the images of two objects
having 0.5, 0.25, and 0.125 nM nanoparticles in each of the objects, respectively.
In all cases, we see that the object(s) are clearly detected. Some artifacts are noted
in the background particularly at the interface between the water background and
the Intralipid solid phantom. Closely examining these images, we observe that the
artifacts exhibit a circular oscillating pattern that is similar to the acoustic pressure
wave distribution (see Figures 7.2a and 7.2b). This may be improved by enlarging
the receiving frequency range, which acts to average out the frequency associated
artifacts. Nonetheless, we see that the optical contrast levels between the object and
background are resolvable for all the cases (see the gray scale in Figures 7.2a to
7.2d). This quantitative nature is further confirmed from the optical property profiles
shown in Figure 7.3.
We next show the ability of the gold nanoparticles to serve as contrast agents for
in vivo tumor imaging with PAT. PEGylated gold nanoparticles (50 nm) (200 μl,
10 mg/mL) were administered via tail vein, and serial photoacoustic images were
obtained following the systemic administration. The accumulation of untargeted
PEGylated nanoparticles inside the tumor following the systemic administration
is expected and has been demonstrated to occur via a process called “enhanced
Contrast Agents–Based Molecular Photoacoustic Tomography 195
30 30 4
(a) 2 (b)
20 20 3.5
3
10 1.5 10
2.5
0 0 2
1
1.5
10 10
0.5 1
20 20
0.5
30 30
30 20 10 0 10 20 30 30 20 10 0 10 20 30
30 30
(c) (d) 1
20 2 20 0.9
0.8
10 1.5 10 0.7
0.6
0 0
1 0.5
0.4
10 10
0.3
0.5 0.2
20 20
0.1
30 30
30 20 10 0 10 20 30 30 20 10 0 10 20 30
FIGURE 7.2 Reconstructed photoacoustic images. (a) A single target containing 0.25 nM
nanoparticles. (b), (c), and (d) Two targets containing 0.5, 0.25, and 0.125 nM nanoparticles in
each target, respectively. The gray scale indicates the relative optical absorption (dimension-
less), while the left and bottom axes refer to spatial coordinates in millimeters.
4.5
4.0 0.50 nM
3.5 0.25 nM
Relative Absorption
0.125 nM
3.0
2.5
2.0
1.5
1.0
0.5
0.0
–10 –8 –6 –4 –2 0 2 4 6 8 10
x or y (mm)
FIGURE 7.3 Optical absorption profiles along a transect through one of the two targets for
the images shown in Figures 7.2b, 7.2c, and 7.2d.
15 15
10 0.2 10 0.2
5 5
x (mm)
x (mm)
0.15 0.15
0 0
5 5
0.1 0.1
10 10
15 0.05 15 0.05
10 0 10 10 0 10
Y (mm) Y (mm)
FIGURE 7.4 PAT images of tumor following tail vein injection of gold nanoparticles at 5
minutes following injection (a) and 5 hours following injection (b). The color scale (right)
represents optical absorption of tissue (arbitrary units). Gold nanoparticles were modified
with PEG5000-thiol following the procedure of Bergen et al. (2006). For these experiments,
human breast cancer cells, BT474 (ATCC), were implanted on the lower abdominal wall of
Nu/Nu mice and allowed to grow to ~1.5 cm in size.
permeability and retention” (Maeda and Matsumura 1989, Kommareddy and Amiji
2007). In the present case, the accumulation of gold nanoparticles inside the tumor
can be visualized by PAT and is shown clearly 5 hours following the systemic admin-
istration of gold nanoparticles (Figure 7.4).
7.2 GRAPHENE NANOSHEETS
Inspired by the strong NIR absorption, high photothermal conversion efficiency, and
exceptionally large surface area of graphene, graphene nanosheets have emerged as
a new high-potential nanomaterial for biomedical applications, especially in the area
of photothermal therapy including photothermal-enhanced drug and gene delivery
systems. It would be highly desirable to monitor/image the in vivo distribution of
multifunctional drug delivery systems, evaluate their posttreatment therapeutic out-
comes in situ, and most importantly to track the long-term fate of graphene sheets in
the human body. These capabilities could largely facilitate their application in practi-
cal multifunctional nanomedicine regimes, fighting various diseases.
Here we describe the use of graphene sheets as contrast agents for PAT and dem-
onstrate the potential of graphene-sheets-based PAT for in vivo monitoring the dis-
tribution of multifunctional drug delivery systems. To generate photoacoustic (PA)
signals with NIR light excitation, the following conditions should be satisfied: strong
NIR absorption, nonradiative relaxation, heating, and acoustic wave generation. We
found that the microwave-enabled low-oxygen graphene (ME-LOGr) nanosheets
exhibit strong and wavelength-independent absorption in the visible and NIR regions
(Figure 7.5). Their absorption (with a coefficient of 22.7 L/g cm at 808 nm) exceeds
the best NIR fluorophores (for example, indocyane green has an absorption coeffi-
cient of 13.9 L/g cm at 808 nm) and the endogenous cellular background. The differ-
ence in NIR absorption between the graphene sheets and the background provides
1.4
5.0 nm ME-LOGr
Nano sheet GO
1.2
1.0
Absorbance (a.u.)
0.8
2.5 nm
0.6
0.4
0.2
250 nm 0.0
0.0 nm
200 400 600 800 1000 1200
Wavelength (nm)
(a) (b)
FIGURE 7.5 (a) AFM images of ME-LOGr nanosheets, (B) UV vs. NIR spectra of
ME-LOGr nanosheets with concentrations of 20 (pink), 10 (olive), 6.7 (blue), 5 (red), and
3.3 mg/L (black), respectively. Inset (b), photograph of an aqueous suspension of ME-LOGr
nanosheets (left) and graphene oxide (GO) nanosheets (right) shows different gray scales,
indicating they are in different oxidation states.
excellent optical confinement for PAT imaging applications. Furthermore, graphene

nanosheets are not luminescent, so that all the optical energy absorbed would trans-
form to heat, which can be used for acoustic wave generation. Therefore, it is reason-
able to assume that a strong NIR PA signal could be generated from these graphene
nanosheets upon NIR illumination.
Figure 7.6 shows that the ME-LOGr nanosheets exhibit remarkably strong PA
signals under NIR laser illumination of 700 nm. In contrast, the graphene oxide
(GO) nanosheets did not show any detectable PA signal at the same concentration
720 nm × 104 800 nm × 104

10 10 2.5
Nano Graphene 3 Nano Graphene
(0.04 mg/ml) (0.04 mg/ml)
2.5 2
5 5
2 1.5
GO GO
0 (0.04 mg/ml) 0 (0.04 mg/ml)
1.5
1
Nano Graphene 1 Nano Graphene
5 5
(0.02 mg/ml) (0.02 mg/ml) 0.5
0.5
10 0 10 0
10 5 0 5 10 10 5 0 5 10
FIGURE 7.6 Photoacoustic images of graphene oxide (GO) and graphene nanosheets of dif-
ferent concentrations, illuminated with 720 and 800 nm pulsed laser coupled with a 1 MHz
central frequency transducer.
and NIR illumination, possibly because of their low NIR absorption capability.
Furthermore, the intensity of the PA signals depends on the concentration of the
ME-LOGr nanosheets, suggesting that the ME-LOGr nanosheets can be used as
NIR contrast agent for in vivo NIR PA imaging. Since the strong NIR absorption of
ME-LOGr nanosheets is almost independent of the wavelength in the NIR region,
their NIR PA signal shows a similar trend of wavelength independence. Figure 7.6
also shows that PA signals generated under 800 nm illumination are similar to those
illuminated at 700 nm. This “wavelength-independent” characteristic is very differ-
ent from other PA contrast agents, such as Au nanorods and Ag nanoplates, which
are highly wavelength dependent (De la Zerda et al. 2011).
7.3 UROKINASE PLASMINOGEN ACTIVATOR RECEPTOR

(UPAR)–TARGETED MAGNETIC IRON OXIDE
NANOPARTICLES (NIR830-ATF-IONP)
Molecular imaging techniques using target-specific probes for positron emission
tomography (PET) or single photon emission computed tomography (SPECT) have
been demonstrated for cancer diagnosis and treatment monitoring (Massoud and
Gambhir 2003). These imaging methods have shown improved specificity and sensi-
tivity in cancer detection, compared with conventional mammography. However, PET
and SPECT have low spatial resolution in determining anatomic location of the tumor.
Additionally, their dynamic and time-resolved imaging ability is limited because of the
long half-life of the radiotracers. PET and SPECT both involve ionization radiation.
In this section we describe the in vivo application of iron oxide nanoparticles
(IONP) as contrast agents for high-resolution PAT. IONP has been widely used to
enhance the image contrast for MRI in animals and has recently been approved for
pilot studies in humans. Amino-terminal fragments of uPA conjugated to iron oxide
nanoparticles (ATF-IONP) have been used successfully in in vivo magnetic reso-
nance imaging of mouse mammary tumors because tumor cells selectively bound
and internalized the ATF-IONP, which provided high contrast for MRI (Yang et
al. 2009). Here we use near-infrared dye labeled amino-terminal fragments of uPA
conjugated to iron oxide nanoparticles (NIR830-ATF-IONP) to specifically bind to
uPAR, a cellular receptor highly expressed in many types of human cancer tissues.
7.3.1 Material and Methods

7.3.1.1 Cell Line
Mouse mammary carcinoma cell line 4T1 was used. Cells were cultured at 37°C and 5%
CO2 in a humidified incubator in Dulbecco’s Modified Eagle’s Medium supplemented
with 10% fetal bovine serum and antibiotics. Cells were harvested at 80% confluence.
7.3.1.2 Preparation of NIR830-ATF-IONP and Control NIR830-

Bovine Serum Albumin-IONP (NIR830-BSA-IONP)
Recombinant amino-terminal fragments (ATF) were generated from pET101/D-
TOPO expression vectors containing a mouse ATF of the receptor binding domain
of uPA cDNA sequence and expressed in E. coli BL21 (Invitrogen, Carlsbad, CA).
ATF peptides were purified from bacterial extracts with Ni2+ nitrilotri-acetic acid
(NTA)-agarose columns (Qiagen, Valencia, CA) using established protocols (Yang
et al. 2009). IR-783 (Sigma-Aldrich, St. Louis, MO) was used to synthesize near-
infrared dye (NIR-830) as described by Lipowska et al. (1993). The schematic and
spectral characterization (excitation wavelength: 800 nm, emission wavelength:
825 nm) of NIR-830 dye were shown in Figures 7.7a and 7.7c. Free thiol groups on
the ATF peptide or control bovine serum albumin protein (BSA) were labeled with
NIR-830 dye and then conjugated to amphiphilic polymer-coated, 10 nm magnetic
iron oxide nanoparticles (IONPs) (Oceannanotech, LLS, Springdale, AR) via cross-
linking of carboxyl groups of the amphilphilic polymer to the animo side groups of
the peptides (Figure 7.7b) using an established protocol. Briefly, cysteine residues of
ATF or BSA were subjected to reduction by TCEP (5 mM at pH 7.4 for 30 mins at
RT). Immediately after reduction, NIR830-maleimide dyes were added and allowed
ATF of uPA
H O
O N
C N
NIR-830
O
S IONP
+
N N
SO–3 SO3Na
(a) (b)
0.15 20
NIR830-ATF
0.12 16
0.09 12
Fluorescence
Absorption
0.06 8
0.03 4
0.00 0
400 500 600 700 800 900
(c)
FIGURE 7.7 Illustration of NIR-830 dye and ATF-conjugated IONP probe. (a) Schematic of
NIR-830 dye. (b) NIR-830 dye is conjugated to mouse ATF peptide through a bond between
maleimide esters and free thiol groups of cystidine residues of the peptide. Dye-labeled
peptides were then conjugated to carboxyl group of the polymer coating on the IONPs.
(c) Resulting optical probe has an Ex 800 nm and Em 825 nm.
for conjugation for 4 hours at room temperature. The amphiphilic polymer-coated

iron oxide nanoparticles (IONPs) were activated with ethyl-3dimethyl amino propyl
carbodiimide (EDAC) and sulfo-N-hydroxysuccinimide (sulfoNHS). The carboxyl
group on the surface of activated IONP was then conjugated to the NIR-830-ATF or
NIR-830-BSA ligands. Unconjugated peptides were removed by washing with 100 k
spin columns three times.
7.3.1.3 Animal Tumor Model

Mouse mammary tumor 4T1 cells (2 × 106) were implanted into mammary fat pads of
6-8-week-old BALB/C mice. Tumors were allowed to grow for 6 to 10 days to a size
of 0.5 to 0.8 cm. Animals were anesthetized with a mixture of ketamine (85 mg/kg)
and xylazine, and were sacrificed using University of Florida Institutional Animal
Care and Use Committee (IACUC)-approved techniques. Strict animal care proce-
dures approved by the University of Florida IACUC and based on guidelines from
the NIH, guide for the Care and Use of Laboratory Animals were followed.
To investigate the feasibility of photoacoustic contrast enhancement with targeted
NIR830-ATF-IONP, we performed in vivo PA imaging of 4T1 mouse mammary
tumors using the following three groups of mice: (1) Group one (n = 3) received an
uPAR targeted IONP targeting agent (100pmol NIR830-ATF-IONP), (2) Group two
(n = 2) received a nontargeted IONP agent conjugated with bovine serum albumin
(BSA) (100pmol NIR830-BSA-IONP), and (3) Group three (n = 2) received no IONP
injection as control. Nanoparticle probes were injected via the tail vein.
7.3.1.4 Photoacoustic Microscopy Imaging System

The schematic of photoacoustic microscopy system was shown in Figure 7.8a. Two
pulsed lasers were used in this study: (1) a tunable Ti:Sapphire laser (LT-2211A,
LOTIS TII) with 8–30 nanosecond (ns) pulse duration and 10Hz repetition rate
for macroscopic imaging of tumors; and (2) a Nd:YAG laser (NL 303HT from
EKSPLA, Lithuania) with 6ns pulse duration and 10Hz repetition rate for micro-
scopic imaging of the blood vessels. The laser beam was split and coupled into
two optical fiber bundles separately, both of which were mounted and adjusted to
allow optimal illumination in the imaging area. Induced photoacoustic waves were
collected by a focused ultrasound transducer (50 MHz or 3.5 MHz). The 50 MHz
transducer with 3 mm aperture and 6 mm focal length yields 30 µm resolution in
axial and 60 µm resolution in lateral at the focal point. The 3.5 MHz transducer
(V383, Olympus) with 15 mm aperture and 35 mm focal length yields axial and
lateral resolution of 400 µm and 820 µm. The imaging probe, transducer, and opti-
cal fiber bundles were mounted on a 2D moving stage. One-dimensional depth-
resolved images (A-line) at each transducer location were acquired and additional
scanning along a transverse direction produced the 2D images, referred to as
B-scans. Further raster scanning along the other transverse direction enabled the
reconstruction of 3D images. All photoacoustic images were displayed in maxi-
mum amplitude projection (MAP) form. All in vivo experiments were performed
with a light intensity of 8 mJ/cm 2, which is lower than the American National
Standards Institute safety limit of 20 mJ/cm 2.
Beam splitter
(a) Beam expander
Pulsed laser
Lens
PC
Lens Fiber
Pre-amplifier Bundle I
Transducer
(b) Fiber Bundle III

Fiber Bundle II
CCD
2D Scanner
Fiber Bundle IV
Filter
Animal
Holder
FIGURE 7.8 Schematic of imaging systems. (a) Schematic of photoacoustic microscopy

imaging system. (b) Schematic of NIR fluorescence imaging system.
7.3.1.5 Near-Infrared Planar Fluorescence Imaging System

As shown in Figure 7.8b, two laser beams from separate 785 nm CW lasers (M5-785-
0080, Thorlabs) were coupled into optical fiber bundle III and optical fiber bundle
IV, which were both fixed and adjusted for homogeneous illumination. A high-
performance fluorescent band-pass filter (NT86-381, Edmund Optics) was mounted
onto the front of a fast charge-coupled device camera (CoolSNAP EZ, Photometrics)
which was used to collect the fluorescence signal. All experiments were conducted
using the same power illumination and camera exposure times.
7.3.1.6 Image Processing
Photoacoustics tomograms were reconstructed by a program implemented in
MATLAB® 7.0 and merged with Amira 5.3.3. The photoacoustic signals collected
were processed by the Hilbert transform before reconstruction, and the photoacous-
tic signals from each mouse were normalized to the same scale (0~256). Fluorescent
images were collected by RS Image (Roper Scientific, Inc) provided by the manufac-
turer and processed by MATLAB 7.0 and fused through Amira 5.3.3.
7.3.1.7 Histological Analysis
Tumors collected from mice were preserved in 10% neutral buffered formalin for
10 hours at room temperature. Histological sections were stained with Prussian blue
staining and analyzed using standard procedures to confirm the presence of iron
oxide nanoparticles.
7.3.1.8 Statistical Analysis
All data obtained from the experiments were summarized using means ± standard
error of the mean (SEM). All data points within a given field-of-view (FOV) (i.e.,
the tumor region in this study) were used to calculate the means/standard errors
presented in this study. For the nontargeted or control tumors, four fiducial markers
were used to indicate the FOV for the analysis.
7.3.2 Results
The absorbance of light by IONPs is much stronger compared with blood at wavelength
longer than 650 nm. The absorbance of IONPs decreases with increasing wavelength
(Figure 7.9b). Although the strongest absorption for IONPs lies near 650 nm, the appropri-
ate wavelength for medical imaging is within the transparent optical window of 700 to
1000 nm, which increases the penetration depth. We measured the photoacoustic signal
of NIR830-ATF-IONP at wavelengths from 730 to 870 nm (Figure 7.9b), and our findings
agreed well with IONP absorption spectra presented by Galanzha et al. (2009). Tumor-
bearing mice were imaged at 24 hours after injection of NIR830-ATF-IONPm and
730 nm was chosen as the best wavelength for in vivo photoacoustic imaging. From the
MAP photoacoustic images shown in Figure 7.9a, NIR830-ATF-IONP-targeted tumors
displayed the highest absorption at 730 nm compared with 800 nm and 850 nm. From the
1
730 nm 800 nm 850 nm
Normalized PA Signal
2 mm
0
(a)
Absorbance/PA Signal (a.u.)
40 NIR830-ATF-IO
IO absorption 4000
35 Blood absorption
3000
30
25 2000
20
15 1000
10
0
650 700 750 800 850 900 730 nm 800 nm 870 nm
Wavelength (nm) Wavelength
(b) (c)
FIGURE 7.9 (a) In vivo MAP photoacoustic images of tumors using signals of 730 nm,
800 nm, and 870 nm, 24 hours after injection of NIR830-ATF-IONP. (b) Absorption char-
acteristics of NIR830-ATF-IONP, IONP, and blood. (c) In vivo PA signals using different
wavelength inputs.
quantitative plot shown in Figure 7.9c, the in vivo photoacoustic signal at 730 nm is 21 and
49% greater, compared with 800 and 870 nm, respectively. Similarly, the absorption of
NIR830-ATF-IONP at 730 nm was 18 and 47% higher, compared with 800 and 870 nm,
respectively. This in vivo data is consistent with in vitro signal enhancement studies.
As shown in Figures 7.10b, 7.10f, and 7.10j, the contrast between tumor and nor-
mal tissue is low before injection. At 24 hours post-NIR830-ATF-IONP injection, the
contrast increased significantly while there was no significant contrast increase using
NIR830-ATF-IO NIR830-BSA-IO Control

(a) (e) (i)
FL image
24 hrs
(b) 2 mm (f) (j)
0 hr
(c) (g) (k)
5 hr
(d) (h) (l)
24 hr
Min Photoacoustic Signal (a.u.) Max
FIGURE 7.10 In vivo photoacoustic MAP and fluorescence images before and after injec-
tion. Micrographs were merged with fluorescence images taken 24 hours postinjection with
indicated agent (a, e, i). (b)–(l) Photoacoustic MAP images were merged with images of blood
vessels before injection (b, f, j), and at 5 hours (c, g, k) and 24 (d, h, l) hours postinjection.
the nontargeted NIR830-BSA-IONP or in animals without injection (Figures 7.10d,

7.10h, and 7.10l). To verify the delivery of NIR830-ATF-IONP to tumor cells,
NIR fluorescent images were collected before each photoaocustic experiment. We
observed the strongest NIR signal in the tumor site at 24 hours postinjection for
NIR830-ATF-IONP (Figure 7.10a), and the strongest NIR signal for NIR830-BSA-
IONP was in the spleen (Figure 7.10e).
We plotted the photoacoustic enhancement in the tumor as a function of time in
Figure 7.11a. The photoacoustic signals with NIR830-ATF-IONP increased 3 and 10
times at 5 and 24 hours postinjection compared with noninjection control mice. This result
indicates that NIR830-ATF-IONP accumulated in the tumors, as expected. Figure 7.11b
shows the quantitative comparisons of photoacoustic signals in the tumor using differ-
ent concentrations of NIR830-ATF-IONP. Compared with noninjection control mice, the
photoacoustic contrast enhancement for 50, 100, and 170 pmol is 300, 1000, and 1200%,
respectively (Figure 7.11b). The trend of photoacoustic signal enhancement over concen-
tration was significantly different. Specifically, the photoacoustic signal increased steeply
from 50 to 100 pmol and then slowly from 100 to 170 pM. Tumors were resected and
imaged by the NIR fluorescent system (Figure 7.11c). The NIR signal inside the tumor
4000 4000
3000 3000
2000 2000
1000 1000
0 0
rs
rs
n
l
ol
ol
ol
ro
tio
ou
ou
BS
pm
pm
pm
nt
jec
H
H
Co
50
0
5
24
In
10
17
e-
Pr
(a) (b)
(c)
FIGURE 7.11 Quantitative plot and comparison of photoacoustic signals. (a) Photoacoustic
signals before and after injection of NIR830-ATF-IONP. (b) Comparison of photoacoustic
signals with different concentrations of injected NIR830-ATF-IONP. (c) NIR fluorescence
imaging of resected tumor (left panel), cross-section of tumor along black dashed line (mid-
dle panel), and section stained with Prussian blue (right panel).
45
2 mm 40
35
SBR (dB)
30
25
20
15
0 10 20 30
Depth (mm)
(a) (b)
PA Signal (a.u.)
0 1 d5 d6
d1 d2 d3 d4
2 mm
31 mm
25 mm
16 mm
12 mm
0 mm 5 mm
(c)
FIGURE 7.12 In vivo photoacoustic images with adding chicken breast between detector
and tumor. (a) MAP image of tumor at 24 hours after injection without adding chicken breast
(left) and photograph of 31mm-thick chicken breast (right). (b) SNR versus depth. (c) B-scan
images by adding different thickness of chicken breast.
is strong and no NIR signals from the surrounding tissue or organs. The right panel of
Figure 7.11c shows the histology section stained with Prussian blue staining. The arrow
indicates the IONP cluster inside the tumor.
To evaluate the clinical utility of this technique, the imaging depth was inves-
tigated by adding biological tissues (chicken breast) to the top of the mice skin.
Figure 7.12a (left) shows the MAP of the tumor in mice that received NIR830-ATF-
IONP at 24 hours postinjection without adding chicken breast. The dashed line repre-
sents the position of B-scan images shown in Figure 7.12c. Even after adding 31 mm
of chicken breast (Figure 7.12a, right), the tumor is clearly seen in Figure 7.12c (d6).
As the imaging depth increased, the signal-to-background ratio (SBR) decreased
from 40 to 18 dB as shown in Figure 7.12b.
7.3.3 Discussion
The photoacoustic signal enhancement for NIR830-ATF-IONP was three times
higher compared to that for nontargeted NIR830-BSA-IONP and 10 times higher
TABLE 7.1
Comparison of Nanoparticles Used as Contrast Agent for Photoacoustic
Tomography
Retention
__Nanoparticles Absortion* Size Time** Dose Multimode
Iron oxide Medium 10nm (sphere) >8 hours pmol PAT and
MRI
Gold nanocage Medium 50nm < 5 hours pmol-nmol NA
Gold nanorod High 52nm × 15nm NA nmol NA
Gold nanoshell High 140nm NA pmol-nmol NA
Gold nantotube Medium 100nm × 11nm NA nmol NA
Single-walled
carbon nanotube Low 186nm × 1nm <2 hours μmol NA
* The wavelength range for the comparison is limited to between 680nm and 800nm.
** The retention time is estimated relative to the size of nanoparticles.
than the noninjection controls. We compared IONP with other nanoparticles used as
contrast agents for PAT in Table 7.1. From the detailed comparison, it is noted that
IONP has modest absorption, smaller size (10 nm) and longer retention time, and
can be used with small dose. These advantages make IONP a good contrast agent for
photoacoustic imaging.
The uPAR-targeted IONP nanoparticle probe used in this study addresses some
of the challenges for the use of targeted tumor-imaging agents. Our methodology uti-
lizes stable and high-affinity targeting ligands and produces strong imaging contrast
for multi-image modalities. It has been shown that human breast cancer and tumor
stromal cells have a higher level of uPAR receptor compared with other normal
breast tissues (DelVecchio et al. 1993, Li et al. 1999). In addition, the highest level
of uPAR receptor expression is detected in the invasive edge of the tumor region
usually enriched in blood vessels making it accessible for uPAR receptor-targeted
IONP in this area (Dublin et al. 2000, Hemsen et al. 2003). The high-quality and
uniformly sized IONPs used in this study were coated with a thin ampiphilic copo-
lymer and have a relatively small particle complex (25 nm), which is more suitable
for in vivo delivery of the imaging probe compared with other receptor-targeted or
nontargeted nanoprobes for photoacoustic imaging (Zerda et al. 2008). It also has
been shown that polymer-coated IONPs have more than 8 hours of plasma retention
time (Moore et al. 2000) compared with other molecular imaging agents for PAT,
which are usually less than 2 hours (Zerda et al. 2008). The delivery of NIR830-
ATF-IONP to the tumor was verified by planar NIR fluorescent imaging. This
shows the potential for us to combine fluorescence molecular tomography (FMT)
with photoacoustic tomography (see Section 7.4 below). We have shown in Chapter 6
that while FMT has lower spatial resolution than PAT, it has higher specificity than
photoacoustic imaging techniques, and the combination of PAT and FMT would
provide complementary information for receptor-targeted molecular imaging.
The demonstrated imaging depth of 31 mm indicates the potential of our method for
tumor imaging in humans. While the photoaocustic system as described here is not yet
suitable for clinical applications, it can be improved to be a clinical prototype using a
commercial ultrasound array and/or a high frame rate of laser pulses with a high-speed
scanning system. By using high-frequency focused ultrasound transducer (>50 MHz), the
microvasculature inside and around the tumor can be imaged as shown in Figure 7.10,
indicating the potential of understanding the interplay between the tumor microvascula-
ture and delivery of targeted contrast agents.
7.4 HER-2/NEU TARGETED MAGNETIC IRON

OXIDE NANOPARTICLES FOR DUAL-MODAL
PHOTOACOUSTIC AND FLUORESCENCE MOLECULAR
TOMOGRAPHY (PAT/FMT) OF OVARIAN CANCER
Ovarian cancer is considered a “silent killer” due to lack of effective approaches
for early detection and therapy. Most ovarian cancer patients are diagnosed at the
advanced stage and don’t respond well to conventional therapeutics. The mortality rate
of the ovarian cancer remains one of the worst among all cancer types (Hoskins 1995).
Ultrasound (US), magnetic resonance imaging (MRI), and computed tomography (CT)
are the most widely used clinical imaging systems for the detection of ovarian cancer.
However, these traditional imaging modalities depend on morphological characteris-
tics and have low specificity and sensitivity in the detection of small tumors (Tammela
and Lele 2004, Shaaban and Rezvani 2009). Over the past several years, there has been
a noteworthy interest in the use of imaging probes that specifically target ovarian can-
cer cells for enhanced sensitivity and specificity in tumor detection. Positron emission
tomography (PET) and single photon emission tomography (SPECT) in combination
with targeted image probes have been used clinically to detect various cancers with
high specificity and sensitivity. However, PET and SPECT have a low resolution to
precisely locate tumor lesions in the peritoneal cavity (Pan et al. 2011, von Schulthess
et al. 2006). Therefore, there is an urgent need to develop noninvasive approaches with
high resolution, sensitivity, and specificity for the detection of ovarian cancer.
We have seen that PAT has characteristics that are well-suited for the devel-
opment of a new molecular imaging technique. Of all the available PAT contrast
agents, iron oxide nanoparticle (IONP) is considered as a potential PAT agent
because of the following properties: (a) modest light absorption in near-infrared
wavelengths, (b) biodegradable and biocompatible with multifunctional charac-
ters, and (c) ability of carrying therapeutic agents for the production of theranostic
nanoparticles for PAT monitoring drug delivery and tumor responses to therapies.
IONP-mediated photoacoustic effect was used for the detection of circulating tumor
cells in the blood of tumor-bearing mice (Galanzha et al. 2009). In Section 7.3, we
have shown the contrast enhancement in breast cancer lesions for PAT imaging
after systemic delivery of a receptor targeted IONPs. Therefore, IONPs have the
potential to be good PAT contrast agents. However, PAT has relatively low sensitiv-
ity compared to pure optical imaging due to the limitations of acoustic transducers
(see Section 6.2). Additionally, the presence of endogenous photoacoustic signals in
some normal organs, such as the kidney, liver, spleen, and heart, may interfere with
specificity of PAT imaging.
Fluorescence molecular tomography (FMT) is an inexpensive and fast 3D optical
imaging modality that has been used for molecular imaging. In contrast to conven-
tional 2D fluorescence molecular imaging, FMT provides improved localization and
quantification in deep tissues (Jiang 2010). In Chapter 6, we have shown that FMT
is more sensitive and specific for the detection of low-concentration contrast agents
in tissues compared with PAT. However, FMT has limited spatial resolution. Hence
the combination of these two modalities with targeted nanoprobes should allow non-
invasive imaging of ovarian tumors with high resolution, specificity, and sensitivity.
Results of a recent study have shown that near-infrared dye labeled ZHER2:342 con-
jugated IONPs (NIR-830-ZHER2:342-IONP) specifically targeted to primary and meta-
static tumors in an orthotopic human ovarian cancer xenograft model and produced
strong imaging signals for 2D-optical and magnetic resonance imaging (Satpathy et
al. 2013). In this study, we wanted to determine the feasibility of application of this
HER-2 targeted IONP as a multimodality imaging probe for the detection of ovarian
cancer using both FMT and PAT.
7.4.1 Methods
7.4.1.1 Cell Lines
Cells from the human ovarian cancer cell line SKOV3, stably expressing a firefly
luciferase gene (SKOV3-luc), were cultured in McCoy’s 5A (Cellgro, Mediatech
Inc. VA, USA) supplemented with 10% fetal bovine serum (Hyclone from Thermo
Scientific) and 1% penicillin streptomycin (Hyclone, Logan, Utah). Cultured cells
were maintained at 37°C in a humidified atmosphere containing 5% CO2.
7.4.1.2 HER-2/neu Specific Affibody Conjugation to IONP

A histidine-tagged HER-2-specific Affibody (His6-ZHER2:342-Cys), was obtained
from the National Institute of Biomedical Imaging and Bioengineering (NIBIB),
Bethesda, Maryland (Lee et al. 2008). ZHER2:342 is a 58 amino acid scaffold protein
with a molecular weight of ~6.5 kDa. The NIR dye, NIR-830 maleimide with excita-
tion and emission spectrum of 797/815 nm (Figure 7.13a) was used for site specific
labeling of the Affibody. Iron oxide nanoparticles with a core size of 10nm were
obtained from Ocean Nano Tech, LLC (Springdale, AR), and coated with amphi-
philic polymers by using established protocols (Yang et al. 2009). The amphiphilic
polymers have carboxyl groups for bioconjugation with amine groups of ZHER-2:342-
NIR-830 (Figure 7.13b). Briefly, the ZHER2:342 which contains a unique C-terminal
cysteine residue for thiol reactive maleimide dye labeling was subjected to reduction
with 5 mM (mmol/L) TCEP for 15 minutes at room temperature. The maleimide
form of NIR-830 dye was then added to ZHER2:342-TCEP mixture (ZHER-2:342: dye
0.14
Photon Intensity (a.u.)
Absorbance
0.12 Fluorescence
0.1
0.08 HER-2 specific Near Infrared dye
0.06 affibody NIR-830
0.04 IO
0.02
IO NIR-830-Z HER2:342-IONP
0
650 700 750 800 850 900
Wavelength (nm) IO nanoparticle
(a) (b)
IONP
0.25
NIR-830-ZHER2:342-IONP
0.20
PA Signal (v)
NIR-830- NIR-830-
BSA-IONP IONP ZHER2:342-IONP 0.15
0.10
0.05
0.00
(c) 650 700 750 800 850 900
Wavelength (nm)
(d)
NIR-830 (4 picomole/uL)
0.06 NIR-830-ZHER2:342-IONP (0.1 picomole/uL) NIR-830-ZHER2:342-IONP
2.5
Ovarian tumor
2.0
0.04
PA Signal (a.u.)
PA Signal
1.5
0.02 1.0
0.5
0.00 0.0
(e) 0 1 2 3 4 5
Concentration (pmol/μL)
(f )
FIGURE 7.13 Schematic and characteristics of imaging probes. (a) Excitation and emis-
sion spectrum of NIR-830 dye. Excitation wavelength: 797 nm; Emission wavelength: 815 nm.
(b) Schematic representation of NIR-830-ZHER2:342-IONP: NIR-830 dye was conjugated to
a cysteine residue at the C-terminal end of ZHER2:342. The ZHER2:342-NIR-830 complex was
conjugated to the carboxyl groups of amphiphilic polymer-coated IONPs (21.6 nm in diam-
eter). (c) Specific binding and uptake efficiency of the NIR-830-ZHER2:342-IONP, IONP, and
NIR-830-BSA-IONP. HER-2 positive SKOV3 cells treated with NIR-830-ZHER2:342-IONP,
IONP, and NIR-830-BSA-IONP were scraped separately and scanned for absorbance from
500 to 784 nm. Since the cells were stained with nuclear fast red, the peak absorbance for
Prussian blue was seen at 528 nm. Absorbance of cells treated with NIR-830-ZHER2:342-
IONP and NIR-830-BSA-IONP at 528 nm was 0.549 and 0.155. (d) Absorbance spectrum
of IONP and NIR-830-ZHER2:342-IONP within NIR range. (e) Photoacoustic evaluation of
NIR-830 and NIR-830-ZHER2:342-IONP with tumors without any contrast agents. The equiva-
lent concentrations of NIR-830 and NIR-830-ZHER2:342-IONP to normal ovarian tumor were
4 pmol/μL and 0.1 pmol/μL. (f) Photoacoustic intensity from different concentrations of
NIR-830-ZHER2:342-IONP.
molar ratio 1:4) overnight at 4°C. Finally, the NIR dye-Affibody conjugated to the
IONPs using ethyl-3-dimethyl amino propyl carbodiimide (EDAC, Sigma-Aldrich)
and N-hydroxy-sulfo-succinimide (sulfo-NHS, Sigma-Aldrich) according to the
carbodiimide method overnight at 4°C (IONP: ZHER-2:342 molar ratio 1:10). Briefly, 100 nmol
of EDAC and 200nmol of sulfo-NHS were added to the aqueous solution of 1 mg of
IONPs, allowing the activation of IONPs for 15 minutes in Borate buffer at pH 5.5.
The reaction was terminated by removing the buffer through a brief centrifugation
using 100 K Nanosep column (Pall Corporation, Port Washington, NY), and then
adding 10mM Borate buffer, pH 8.5. After a brief wash with Borate buffer, pH 8.5,
the reaction of IONPs with ZHER2:342-NIR-830 dye was carried out for 4 to 6 hours
at room temperature. The final product NIR-830-ZHER2:342-IONP was made after
washing the mixture through Nanosep 100 K column three times and conjugate was
stored in 10 mM Borate buffer at pH 8.5.
7.4.1.3 Preparation of NIR-Bovine Serum Albumin-

IONP (NIR-830-BSA-IONPs)
Bovine serum albumin (BSA) was used to produce control nontargeted IONPs.
Briefly, 1mg of BSA (Sigma), at the molar ratio of 1:4, was added to 60 nmol of
NIR-830-maleimide dye and allowed for conjugation for 4 hours at room tempera-
ture or overnight at 4°C. After washing the BSA-dye complex with a 3 K spin col-
umn, the complex was allowed for conjugation with activated IONPs at the molar
ratio of 1:10 (IONP: BSA). The final NIR-830-BSA-IONPs conjugate were purified
using Nanosep 100K column filtration. The hydrodynamic sizes of IONP, NIR-830-
ZHER2:342-IONP and NIR-830-BSA–IONP used for the experiments were measured
by dynamic light scattering (DLS) instrument (Zeta-sizer Nano: ZS90, Malvern
Instruments Ltd.).
7.4.1.4 Evaluation of Labeling Efficiency by Prussian Blue Staining

Approximately 105 of SKOV3 cells were seeded in each cell culture chamber over-
night; 100nM of NIR-830-ZHER2:342-IONP, NIR-830-BSA–IONP or IONPs were added
into the RPMI1640 containing 2% FBS culture media and then incubated at 37°C
for 2 to 3 hours. The medium containing particles were then removed and cells were
washed with PBS three times times followed by fixation with 4% paraformaldehyde
for 20 minutes. After a brief wash with PBS, Prussian blue staining solution prepared
by 1:1 mixture of 5% potassium ferrocyanide and 5% HCl acid was added for 30 min-
utes to 2 hours at 37°C to confirm the presence of intracellular IONPs under optical
microscope (400X magnification). Finally the percentage of iron positive cells was
obtained by counting ~200 cells. Every time when the fresh conjugates were prepared,
specificity of the conjugates was tested in different passages of the SKOV3-luc cells.
7.4.1.5 Orthotopic Human Ovarian Cancer Xenograft Model

The ovarian tumor model was established by injecting 5 × 104 of SKOV3 luciferase
gene positive human ovarian cancer cells into the ovaries of 6 to 8-week-old female
athymic nude mice (Harlan). Mice were imaged when orthotopically xenografted
tumors reached around 5mm in size, usually 4 to 6 weeks after the tumor cell injection.
Beam
Beam splitter
Probe expander
Pulsed laser
Detectors
Sources
CCD Lens
Lens Fiber
Detector PC Pre-amplifier
Bundle I
interface
Source Transducer
Laser Fiber Bundle II
interface
Lens 2D Scanner
Probe
Holder
Animal
Holder
2D linear stage
(a) (b)
FIGURE 7.14 (a) Schematic of the hand-held FMT imaging system. (b) Schematic of the
PAT imaging system.
7.4.1.6 Fluorescence Molecular Tomography Imaging System

A handheld optical fiber array–based FMT system is shown in Figure 7.14a. In this
system, a continuous-wave (CW) 785-nm laser was mounted in a 2D linear stage. A
convex lens was used to couple the laser beam into each source fiber in the array. For
each source illumination, a 1024 × 1024 pixel CCD camera (Princeton Instruments,
Trenton NJ) equipped with a band-pass filter was used to collect the emitted light from
the detection interface. Optimal exposure time was selected for each experiment and
binning of 4 × 4 pixels was used to improve the SNR. Fluorescence images were recon-
structed using an iterative finite element-based algorithm, described in Jiang (2010).
7.4.1.7 Photoacoustic Imaging System

A schematic of the photoacoustic imaging system is shown in Figure 7.14b. A tun-
able Ti:Sapphire laser (LT-2211A, LOTIS TII) with 8 to 30 ns pulse duration and 10 Hz
repetition rate was used. The laser beam was expended, split, and coupled into two
optical fiber bundles adjusted to allow optimal illumination within the imaging area
(20 × 20 mm2). The measured photon energy density on the tissue surface was 8
mJ/cm2. A 3.5 MHz focused ultrasound transducer (V383, Olympus) with 15 mm
aperture and 35 mm focal length was used to receive laser-induced photoacoustic
waves, yielding axial and lateral resolutions of 400 and 820 μm, respectively. The
imaging probe, consisting of transducer and optical fiber bundles, was fixed onto a
2D moving stage. One-dimensional depth-resolved images (A-lines) at each trans-
ducer location were acquired through raster scanning along one transverse direction,
and formed 2D images (B-scans). Further raster scanning along the other trans-
verse direction enabled the formation of 3D images. The total raster scanning points
was 100 A-lines along each direction with an interval of 200 μm over a distance
of 20 mm. It took 20 minutes for one scan, and most 3D PAT images are shown in
maximum amplitude projection (MAP) form.
7.4.1.8 In Vivo Planar Fluorescence Imaging

Planar fluorescence imaging was performed 24 and 48 hours after systemic delivery
of NIR-830-ZHER2:342-IONP using a Kodak in vivo imaging system FX (Carestream
Health Inc., New Haven, CT) to demonstrate specific accumulation of the targeted
IONPs in the ovarian tumors, and anatomically localize the tumors. Briefly, an
800nm excitation filter and an 850-nm emission filter were used. The imaging sys-
tem also includes F-stop 2.5 and 100mm field of view (FOV). The in vivo images
were captured by an excitation time of 3 min with a gamma value of 0.2. By using
the built-in Kodak software, the mean fluorescence intensity (MFI) was measured
over a region of interest (ROI) in the tumor area and the surrounding skin area,
and the corresponding ratio was calculated as the signal-to-body background ratio
(SBR).
7.4.1.9 In Vivo Imaging in Animal Tumor Models

Bioluminescence images were carried out weekly after the injection of 30 mg/kg
luciferin substrate in the anesthetized mice to track the SKOV3-luc orthotopic tumor
growth using the IVIS Imaging System (Xenogen) (see Figure 7.15). Acquisition
times started with 1 minute and were reduced later on to avoid signal saturation.
Signal strength was quantified with LIVINGIMAGE software (Xenogen) by mea-
suring photon flux over a ROI.
Three groups of mice were examined using both FMT and PAT. The first group (n = 4)
received an injection of 400 pmol of NIR-830-ZHER2:342-IONP. The second group (n = 3)
received an injection of 400 pmol of NIR-830-BSA-IONP. The third group (n = 3) without
injection was used as a control.
7.4.1.10 Histology Analysis
Tumors along with normal organs were collected after sacrificing the mice. The tis-
sues were fixed with 10% buffered formalin. Paraffin tissue sections were stained
with Prussian blue or hematoxylin and eosin (H&E). Images were acquired at 200X
magnification using a Zeiss Axioplan 2 upright microscope.
6 ×105
5 ×105
4 ×105
3 ×105
2 ×105
1 ×105
Week 2 Week 3 Week 4 Week 5 Week 6
Counts (p/s)
FIGURE 7.15 Bioluminescence images of a representative mouse following implantation of

the SKOV3-luc cells in the left ovary.
7.4.1.11 Statistical Analysis
All data used for statistical analysis were summarized using means ± standard error
of the mean (SEM).
7.4.2 Results
Taking advantage of smaller size and highly affinity nature of the Affibody, we have
used a relatively new class of affinity molecule, HER-2 Affibody (ZHER2:342), for conju-
gating to IONPs. Prior to the conjugation, ZHER2:342 was prelabeled with a near-infrared
dye (NIR-830 dye) to the cysteine residue of the Affibody molecule. The biconjugate
NIR-830-ZHER2:342 was finally coupled to the carboxyl group of the amphiphilic poly-
mer coating of IONP. Figure 7.13b depicts the schematic representation of the pro-
duction of the imaging probe, NIR-830-ZHER2:342-IONP. By using Zeta-sizer Nano,
the hydrodynamic size of the IONP was 14.7 ± 3.58nm and increased upon conju-
gating with ZHER2:342 (NIR-830-ZHER2:342-IONP: 21.6 ± 5.61nm) and BSA (NIR-830-
BSA–IONP: 26.12 ± 2.22nm). Specific binding and uptake efficiency of the targeted
nanoparticles (NIR-830-ZHER2:342-IONP) were examined in the HER-2 over expressing
SKOV3 cells as shown in Figure 7.13c. In general, when 100nM of Fe equivalent con-
centration of targeted IONPs were incubated with SKOV3 cells for 5 hours followed
by 1 hour Prussian blue stain, about 60% cells were blue-stained iron-positive cells.
The iron-stained cells increased with higher concentrations of targeted nanoparticles.
However, cellular uptake by nontargeted IONPs was significantly lower when SKOV3
cells were incubated with nontargeted IONPs under similar conditions (Figure 7.13c).
The absorbance spectrum of IONPs and NIR-830-ZHER2:342-IONP used for photo-
acoustic imaging was evaluated within NIR range and shown in Figure 7.13d, where
absorption of IONPs decreased as the wavelength increased. From the spectrum, we
noted that even the absorption of NIR-830-ZHER2:342-IONP had a peak value at 820
nm contributed by the NIR dye, its absorption at 730 nm was still higher than that
at 820nm since IONPs were likely to be the major contributor to the photoacoustic
enhancement due to a low concentration of NIR-830 dye molecules on each IONP
(10 dye molecules per IONP). To verify the assumption, we compared the contribu-
tion to the photoacoustic effect by pure NIR-830 dye and NIR-830-ZHER2:342-IONP.
From the results shown in Figure 7.13e, we noticed that when the concentration of
NIR-830 dye was 40 times higher than the concentration of IONPs in NIR-830-
ZHER2:342-IONP, they contributed equally to the photoacoustic effects. That means
one IONP equals 30 dye molecules in the generation of photoacoustic signals with
consideration of 10 dye molecules per IONP in NIR-830-ZHER2:342-IONP.
To obtain the equivalent concentration of IONPs in NIR-830-ZHER2:342-IONP to nor-
mal ovarian tumors in photoacoustic imaging, we compared the photoacoustic signals
of different concentrations of IONPs in NIR-830-ZHER2:342-IONP with that of the tumors
without any contrast agents. Photoacoustic signals decreased as expected when the
concentration of IONPs in NIR-830-ZHER2:342-IONP decreased. When the concentra-
tion of IONPs reduced to 0.1 pmol/μL, the amplitude of photoacoustic signals becomes
almost the same as that from the tumor (Figure 7.13e). The amplitude change of pho-
toacoustic signals relevant to the concentrations of IONPs in NIR-830-ZHER2:342-IONP
was quantitatively plotted in Figure 7.13e. A linear regression fit of the data yielded an
NIR 830 dye-ZHER2:342-IONP

(pmol/uL)
0.9 NIR-830-ZHER2:342-IONPs
0.9 pmol/uL 0.18 pmol/uL
× 104
10 10
0.18 6
5
0 0 × 105
5 6
5
0.09 10 10
4 4
10 0 10 10 0 10
3 2
0.06 0.04 pmol/uL 0.04 pmol/uL
0
2
10 10
5 5
1
0.04 0 0
0 5 5
10 10
10 0 10 10 0 10
0.01 (b)
150 ms
(a)
FIGURE 7.16 In vitro evaluation of NIR-830-ZHER2:342-IONPs using planar fluorescence

imaging (a) and FMT (b). As we expected, the optical signals and reconstructed fluorescence
intensity decreased when the concentration decreased. When the concentration of NIR-830-
ZHER2:342-IONPs was lower than 0.04 pmol/μL, both planar fluorescence imaging and FMT
still can detect it. When the concentration was reduced to be 0.02 picomole/μL, either planar
fluorescence imaging or FMT cannot image it.
R2 equaling to 0.99, which was expected given the amplitude of the photoacoustic sig-
nal was proportional to the concentration of IONPs in NIR-830-ZHER2:342-IONP.
To compare the sensitivity of FMT and PAT in imaging when NIR-830-ZHER2:342-
IONP produced contrast, we performed planar fluorescence and FMT phantom
experiments using different concentrations of IONPs in NIR-830-ZHER2:342-IONP. We
found that when the concentration reduced to 0.04 pmol/μL, they were still detect-
able for both planar fluorescence imaging and FMT (Figure 7.16). In the case of
photoacoustic imaging, there was no enhancement when the concentration became
as low as 0.1 pmol/μL. Results of our study suggested that this handheld FMT has
higher sensitivity compared to PAT for imaging NIR-830-ZHER2:342-IONP.
From in vivo planar fluorescence imaging, it showed that the fluorescence inten-
sity reached to the strongest level in the ovarian cancer at 48 hours following sys-
temic delivery (Figure 7.17) when the tumor size was <5 mm (Satpathy et al. 2013).
However, in tumor-bearing mice without receiving injection of the imaging probes
or injected with NIR-830-BSA-IONP, no significant imaging signals were detected
969
24h 48h
772
Tumor
region
575
377
180
FIGURE 7.17 HER-2 targeted imaging of orthotopic ovarian tumors following systemic
delivery of NIR-830-ZHER2:342-IONP detected by 2D-NIR optical imaging. Optical imaging
was performed 24 and 48 hours postinjection using the Kodak FX in vivo imaging system. The
NIR signal scale was generated using the Kodak molecular imaging software. Anatomical
x-ray images were simultaneously captured to validate tissue position.
in the tumor region. The mean fluorescence intensity (MFI) of the tumor for the
animal groups that received injection of NIR-830-ZHER2:342-IONP or NIR-830-BSA-
IONP 24 and 48 hours postinjection are plotted in Figure 7.18a. The high level of
the MFI found in the tumors of the mice that received NIR-830-ZHER2:342-IONP sug-
gested effective accumulation of the nanoprobes in the tumor at both time points.
Four time points were selected to perform photoacoustic imaging, and quantita-
tive analysis was shown in Figure 7.18b. Photoacoustic contrasts in tumors of mice
before injection of imaging probes were comparable. Twenty-four hours postinjec-
tion, the contrast of tumors of mice that received NIR-830-ZHER2:342-IONP increased
significantly and reached the peak value 48 hours postinjection, which is consistent
with the planar fluorescence imaging. However, no significant contrast enhancement
appeared in nontargeted and noninjection groups.
All three groups of mice were imaged by both PAT and FMT. The imaging area
was guided by both bioluminescence imaging (BLI) (Figure 7.18c) and planar fluo-
rescence imaging (Figure 7.18d). As expected, tumors of the mice that received injec-
tion of NIR-830-ZHER2:342-IONP had the strongest photoacoustic signal (left image
in Figure 7.18e), compared with the tumors of the mice that received injection of
NIR-830-BSA-IONP (middle image in Figure 7.18e) or mice in the no injection con-
trol group (right image in Figure 7.18e). The corresponding FMT results were shown
in Figure 7.18f, which was consistent with the results of planar fluorescence imag-
ing and PAT. Data from multispectral plots shown in Figure 7.18g revealed that the
strongest photoacoustic signal in the tumors of mice that received injection of NIR-
830-ZHER2:342-IONP was at 730nm, which was 500% higher than that of noninjection
controls, and 300% greater than that of the mice injected with NIR-830-BSA-IONP.
After the mice were sacrificed, tumors as well as normal organs such as the liver,
kidney, spleen, and heart in targeted and nontargeted groups were resected and eval-
uated by planar fluorescence imaging, PAT and FMT. In fluorescence imaging, we
found strong optical signals in the kidney and liver but not in other normal organs
(Figures 7.19a and 7.19b). However, in PAT images, the tumor of targeted group
2.50 Targeted
N=4 Normal tumor
P<0.005*** P<0.005*** 4
2.00
PA Signal (a.u.)
Non-targeted
1.50 3
MFT
1.00 2
0.50 1
0.00 0
24 h 48 h 24 h 48 h 0 20 40 60
NIR-830-ZHER2:342-IONPNIR-830-BSA-IONP Hours
(a) (b)
Non-targeted Non-targeted
Targeted Control Targeted Control
4
(c) 6 ×10 (d) 6.5 ×104
1 ×104 3.5 ×104
Targeted Non-targeted Control Targeted Non-targeted Control

(e) 1 (f) 1
MAP
FMT (a.u.)
PAT (a.u.)
MAP
1 mm
B-scan Surface Cross Surface

0 section 0
Control Mice with

Mice with non-target injection
4
targeted injection
3
1
0
nm
nm
nm
0
0
73
80
87
Wavelength (nm)
(g)
FIGURE 7.18 In vivo fluorescence and photoacoustic imaging of tumor-bearing mice and
quantitative analysis. (a) The mean fluorescence intensity plot of tumors in targeted or non-
targeted group 24 and 48 hours postinjection. Fold increases shown in the bar figure were the
mean fluorescence intensity of the tumor area divided by the mean fluorescence intensity of
the body background. Results were from four mice in each group. Student’s t-test was used to
determine the p value. (b) Bioluminescence imaging of the mice showed the position of the
tumor that was used to select the imaging areas of FMT and PAT. (c) Bioluminescence images
of typical mice from targeted (left), nontargeted (middle), and control (right) animal groups.
(d) Planar fluorescence imaging of mice from targeted, nontargeted, and control groups 24
hours postinjection of nanoparticles. (e, f) MAP images and typical cross-sections of PAT
and FMT of mice in the targeted (left), nontargeted (middle), and noninjection (right) animal
groups. Tumors of the mice from the targeted group were clearly detected by both PAT and
FMT, while the imaging contrasts of the tumors of the mice from nontargeted and noninjec-
tion groups were too low to be detected. (g) Quantitative photoacoustic intensity plots of
tumors from the targeted (red), nontargeted (black) and noninjection (blue) animal groups at
730, 800, and 870 nm, respectively.
Tumor Heart Spleen Kidney Liver

(a)
1 cm
3000
4000
5000
6000
7000
8000
9000
10000
11000
Tumor Heart Spleen Kidney Liver

(b)
1 cm
3000
4000
5000
6000
7000
8000
9000
10000
11000
NIR-830-BSA-IONP
Tumor Liver Kidney Heart Spleen 1 × 104
(c) Planar FL
2D FL
1 mm
0
1
PAT: MAP
PAT (a.u.)
0
1
FMT: MAP
FMT (a.u.)
Tumor Liver Kidney Heart Spleen 1 ×106

(d) Planar FL
2D FL
1 mm
0
1
PAT: MAP
PAT (a.u.)
0
1
FMT (a.u.)
FMT: MAP
FIGURE 7.19 Ex vivo evaluation of tumors and normal organs using planar fluorescence
imaging, PAT, and FMT. Planar fluorescence imaging of the tumor, heart, spleen, kidney, and
liver of the mouse in the targeted (a) and nontargeted (b) groups. PAT and FMT evaluation of
the tumor and small samples of normal organs of the mice in targeted animal group (c) and
nontargeted group (d).
Tumor Tumor Tumor Tumor

PB HF PB HF
Liver Kidney Liver Kiney

PB PB PB PB
Heart Spleen Heart Spleen

PB PB PB PB
NIR-830-ZHER2:342-IONP NIR-830-BSA-IONP
(a)
0.8
NIR-830-BSA-IONP
0.6
0.4
0.2
0.0
or
ey
rt
n
r
ve
e
ea
m
dn
le
Li
Sp
Tu
Ki
(b)
FIGURE 7.20 Histological and quantitative analysis of tumors and normal organs. (a) Prussian
blue staining of tissue sections showed high levels of iron-positive cells in the tumor of mice
in targeted group and low to intermediate levels of iron-positive cells in the livers and spleens
of mice in both targeted and nontargeted groups. No distinct iron-positive cells were visible in
the tumors of mice in nontargeted group, and in the kidney and heart. PB: Prussian blue, HE:
hematoxylin and eosin. (b) Comparison of photoacoustic intensity of the tumors and normal
organs resected from the mice in targeted and nontargeted groups. The heart and targeted
tumor produced the strongest photoacoustic emission. The spleen generated intermediate
photoacoustic emission.
and hearts emitted strongest photoacoustic signals and the photoacoustic signals of
the spleen were stronger than that of the liver and kidney. From histological sec-
tions and quantitative analysis in Figure 7.20, systemic accumulation of IONPs in the
tumor led to strongest photoacoustic emission, and nonspecific uptake of the IONP
by the reticuloendothelial system (RES) in the liver and spleen resulted in a higher
photoacoustic emission compared with that of the kidney. A large volume of blood
in the heart generated stronger photoacoustic signals than other normal organs since
the light absorption of blood at 730nm is stronger than that of IONPs. Additionally,
we did see strong optical signals of the kidney but didn’t detect blue iron-stained
cells in the tissue section of the kidney. This may be caused by the renal clearance of
free dye molecules or breakdown products of targeting ligands in the kidney to retain
the nanoparticles in the tissue section (Satpathy et al. 2013).
To evaluate the clinical potential of PAT/FMT technique for intraoperative imag-
ing of ovarian tumors in humans, we examined the ability of imaging of tumors
located in deep tissues by adding different thicknesses of chicken breast tissues on
the top of mouse back. The top row in Figure 7.21a shows the B-scan PAT images
with the different thicknesses of the chicken breast. The bottom row (Figure 7.21a)
shows the corresponding MAP images. As shown, ovarian tumors of the mice that
received injection of NIR-830-ZHER2:342-IONP were clearly imageable even after the
addition of a lay of 19 mm exogenous tissues. Signal-to-background ratio (SBR)
decreased from 38 to 17 dB when imaging depth increased. We also investigated the
imaging depth of FMT (third and fourth rows in Figure 7.21b) and found an imaging
depth of up to 10 mm.
We then compared the spatial resolution of FMT and PAT imaging at different
depths. Representative 3D images of PAT and FMT with different thicknesses of the
chicken breast are shown in Figure 7.22. MAP images and typical cross-sections of
both PAT and FMT areshown in Figure 7.21b. We found that the axial and lateral
resolutions of FMT were relatively poor. Although the position of image center was
in the right depth, the axial and lateral sizes were much larger than the actual sizes.
In contrast, high lateral and axial resolutions of PAT made the reconstructed lateral
and axial sizes very close to the actual dimensions of the tumor. The recovered
lateral and axial sizes of PAT and FMT were plotted (Figure 7.23) and the average
lateral and axial full-width at half-maximum (FWHM) of the tumor in PAT images
at three different depths (3, 6, and 10mm) were 5 and 3 mm, which are comparable
to the size of the resected tumor. Using FMT, the lateral FWHM was 5 to 13 mm and
axial FWHM was 3 to 14 mm, as the depth increased, which were relatively larger
than the actual tumor size.
7.4.3 Discussion
In this section we demonstrated the feasibility of using PAT and FMT to detect
orthotropic ovarian tumors in mice after systemic delivery of HER-2/neu targeted
imaging probes. In Section 7.3, we compared the PAT imaging signals of IONPs
with several other nanoparticle contrasts, such as carbon nanotubes, gold nano-
cages, gold nanorods, and gold nanoshell. Although our results showed that gold
nanoparticles and carbon nanotubes had stronger photoacoustic contrasts than that
of IONPs in animal tumor models, the effects of systemic delivery of those nanopar-
ticles in humans have yet to be determined since the mechanisms of clearance of the
nanoparticles and short- and long-term toxicity are still unclear. In contrast, IONPs
are a class of biodegradable and biocompatible nanoparticles with a low toxicity in
0 mm 10 mm 19 mm
Surface Surface
Surface
B-scan
MAP
(a)
3 mm 6 mm 10 mm
PAT: MAP
Surface Surface
Surface
PAT: Cross-
section
FMT: MAP
Surface FMT: Cross-

section
(b)
FIGURE 7.21 Evaluation of imaging depth and spatial resolution of PAT and FMT.
(a) B-scan (top row) and MAP (bottom row) PA images of a tumor in a representative mouse
that received NIR-830-ZHER2:342-IONP. The back flank of the mouse was overlaid with 0, 10,
and 19 mm-thick layers of chicken breast to mimic the normal tissue in humans. Red arrows
showed the surfaces of the chicken breast. Scale bar: 5 mm. (b) Dual modality imaging of the
same tumor located at different depth (3, 6, and 10 mm). The dashed white lines indicate the
position of the selected cross-sections of PAT and FMT, and arrows showed the positions of
the imaging surfaces. Scale bar: 5 mm. From the MAP and cross-sections of PAT and FMT,
we noticed that the lateral and axial resolutions of PAT were higher than that of FMT.
3 mm 6 mm 10 mm
PAT (3D)
z
y
x
FMT (3D)
z
y
x
FIGURE 7.22 Dual-modality 3D imaging of the same actively targeted tumor located in
different depth. The back flank of the mouse was overlaid with 3, 6, and 10 mm-thick layers of
chicken breast to mimic the thick normal tissue in humans. The white arrow showed the sur-
faces of the chicken breast tissue in the PAT images. The top surfaces of the boundary boxes
in FMT represented the surfaces of the chicken breast tissue. From the results, we noticed
that the reconstructed lateral and axial sizes of FMT were overestimated when imaging depth
increased, while the reconstructed lateral and axial sizes of PAT were within the range of the
measured size of the tumor.
humans and have already been used in human patients for MRI detection of liver
cancer and lymph node metastasis after systemic delivery (Harisinghani et al. 2003).
Based on our in vitro test result, we found that the photoacoustic signal was gen-
erated primarily from the IONPs as the concentration of NIR-830 dye conjugated
to the IONPs is too low to show a marked signal enhancement effect. However, it is
feasible to enhance the PAT signal by modifying our nanoparticle design to incorpo-
rate large amount of NIR-830-dye molecules (over 1000 molecules per IONP) into
the polymer-coating of the IONPs. Increased concentration of NIR dye in NIR-830-
ZHER2:342-IONP will further enhance the PAT signal.
Planar fluorescence imaging has been commonly used in molecular imaging for
monitoring the delivery of nanoprobes. However, it lacks the ability to map tumors
with the imaging probes that localize deep inside the body because of a poor spatial
resolution and lack of depth information. In this study, the handheld FMT system
showed the potential to noninvasively map ovarian tumors in 3D at depths up to
10 mm. The imaging procedure is fast, taking less than 2 minutes to cover a
20 × 20 mm2 imaging area. Additionally, FMT can specifically detect tumors with
a high sensitivity because of inherent characteristic of fluorescent imaging modality.
However, relatively low axial and lateral resolutions affect the imaging accuracy. As
a complementary method to FMT, PAT was employed to detect ovarian tumors with
contrast enhancements from IONPs. We anticipate that in a clinical setting, the fast
3 mm 3 mm
1 6 mm 1 6 mm
Photoacoustic Signal (a.u.)
Fluorescence Signal (a.u.)

10 mm
10 mm
0.8 0.8
0.6 0.6
0.4 0.4
0.2 0.2
0 0
0 3 6 9 12 15 18 0 3 6 9 12 15 18
x-axis (mm) x-axis (mm)
(a) (b)
3 mm 3 mm
1 6 mm 1 6 mm
Photoacoustic Signal (a.u.)
Fluorescence Signal (a.u.)

10 mm 10 mm
0.8 0.8
0.6 0.6
0.4 0.4
0.2 0.2
0 0
0 3 6 9 12 15 18 0 3 6 9 12 15 18
z-axis (mm) z-axial (mm)
(c) (d)
FIGURE 7.23 Quantitative comparison of spatial resolutions between PAT and FMT with
increased imaging depth. Lateral profiles of the recovered tumor by PAT (a) and FMT
(b) along the dashed white lines in Figure 7.21b. Axial profiles of the recovered tumor by PAT
(c) and FMT (d).
hand-held FMT system could be used to mark the possible tumor-containing areas,
and then PAT could be used to image the marked area to accurately identify tumor
margins and disseminated small tumor lesions in the peritoneal cavity. This combi-
nation imaging approach takes the advantage of both FMT and PAT to complement
each single modality.
8 Clinical Applications
and Animal Studies
In this chapter, we discuss the applications of photoacoustic tomography (PAT) to
several clinically significant areas and demonstrate its validity in humans or in ani-
mals. As described in Chapter 2, quantitative PAT (qPAT) can provide high-resolu-
tion functional images including hemoglobin and oxygenation, which promises its
potential in breast cancer detection. Quantitative PAT is also uniquely suited for
imaging joint diseases (e.g., osteoarthritis). The possibility of miniaturizing a PAT
imaging probe renders it an excellent modality for intravascular, endoscopic, or
intraoperative imaging. The high spatial and temporal resolutions of PAT in microm-
eter and millisecond scales offer its bright future in brain imaging, as it will play an
unprecedented role in noninvasively mapping brain activity and connectivity, one of
the hottest topics in today’s neuroscience and neurology.
8.1 JOINT IMAGING
Osteoarthritis (OA), or degenerative joint disease, is the most common form of joint dis-
ease. It is a slowly progressive disease and a major cause of morbidity in the population
over age 50, affecting more than 21 million Americans. It also imposes considerable
expense on the health care system. Although a number of factors contribute to its devel-
opment, including obesity, trauma, and genetic predisposition, the hallmark of OA is
progressive damage to articular cartilage. As damage progresses, fissuring of the articu-
lar cartilage can result in fragmentation and deposition of small loose bodies within the
joint space, causing a low-grade inflammatory reaction and increased turbidity in the
synovial fluid, with a decrease in viscoelastic and lubricating properties. Subchondral
bone also becomes exposed and results in sclerosis. While there is currently no cure for
this disease, numerous studies have shown that the progression of articular damage may
be modified by medical or surgical intervention if the disease is detected early (Itay et
al. 1987, Howell and Altman 1993, Brittberg et al. 1994). These studies, coupled with
recent developments in gene therapy and anticipated novel therapeutic approaches, have
generated substantial demand for noninvasive techniques for detecting early changes in
the joints, when intervention is likely to have its greatest effect.
Several noninvasive imaging modalities have been applied to detect OA, includ-
ing X-ray radiography, computed tomography (CT), ultrasound (US), and magnetic
resonance imaging (MRI) (Brandt 1997, Chan et al. 1991, Myers et al. 1995, Tyler
et al. 1995). Radiography and CT are excellent techniques to image hard tissues and
are primarily used for advanced OA stages when bone damage has occurred (Brandt
1997, Chan et al. 1991). It is difficult for these techniques to detect changes in carti-
lage or the early inflammatory process in synovial membrane/fluid. Ultrasound has
223
limited utility in evaluation of the early stages of OA and is very operator depen-
dent, making it difficult to be used as a quantitative tool. MRI represents a power-
ful modality for evaluating joint abnormalities, particularly when a surface coil or
contrast agent is used (Tyler et al. 1995, Ostergaard et al. 1997). However, MRI is
expensive and may not be suitable for long-term, routine monitoring of OA.
It has been recently recognized that optical contrast is highly sensitive to structural
and metabolic changes associated with an OA joint (Yuan et al. 2008, Jiang 2010) due
to a strong correlation between optical changes and the diseased joint. Collagen fibril
network deterioration has been observed when OA was present in the cartilage area
(Panula et al. 1998). This deterioration decreased the collagen-induced optical path
difference by 19% to 71%. The variation in fibrils together with the loss of proteogly-
cans in OA should result in significant changes in both the scattering and absorption
coefficients of the cartilage. In addition, it is known that with the onset of OA, the
synovial membrane/fluid in articular cavity becomes increasingly turbid (Beuthan
et al. 1996, Prapavat et al. 1997). The increased turbidity would accompany increased
scattering and absorption coefficients in the diseased synovial membrane/fluid. In
fact, this optical increase can be as large as 100% at certain wavelengths in the near-
infrared (NIR) region (Beuthan et al. 1996, Prapavat et al. 1997), which would provide
good optical contrast for imaging purposes. Finally, there is increasing evidence that
OA is a disease involving a metabolic dysfunction of bone (Mansell 2007, Knott and
Bailey 1998). It is likely that this metabolic dysfunction of bone, often associated with
high metabolism of subchondral bone, will cause changes in its tissue oxygenation,
Hb, HbO2, and water content. These metabolic and functional parameters should also
be measurable by multispectral optical/photoacoustic imaging.
The demonstrated optical detection of OA, however, was based on diffuse optical
tomography (DOT) (Jiang 2010), a pure optical imaging method that has relatively low
spatial resolution (2~4 mm) due to strong tissue scattering. While it is known that an
adult articular cartilage has a typical thickness of from 2 to 4 mm, diseased cartilage
typically has a thickness of 1 mm or smaller (Yuan et al. 2008). Thus, to correctly image
an OA joint, DOT needs to be combined with a high-resolution modality such as X-ray
so that the high-resolution X-ray images can be used as a priori structural guidance
for DOT reconstruction (Yuan et al. 2008). Laser-induced photoacoustic tomography
(PAT) alone can achieve the same as the combined X-ray/DOT does, since it retains
the desired high optical contrast while providing much better spatial resolution than
pure optical tomographic methods by detecting much less-scattering ultrasonic waves.
In addition, PAT is free of the speckle artifacts present in pulse-echo ultrasonography.
Here we demonstrate the application of single- and multispectral quantitative
PAT for imaging OA in the finger joints in both 2D and 3D.
8.1.1 2D Single-Spectral Quantitative PAT

8.1.1.1 Materials and Methods
8.1.1.1.1 Photoacoustic Imager
Our PAT system consists of a pulsed NIR light source, a spherical scanning subsys-
tem, and an ultrasound detection array and associated acoustic signal/data acqui-
sition. In this system, pulsed light beam at 805 nm, generated by a Ti:Sapphire
Clinical Applications and Animal Studies 225
(a) (b)
T7 T6
T8
T5
T4
T3
T2
T1
Section
line
FIGURE 8.1 (a) Photograph of the transducer array/finger interface of the photoacous-
tic tomography system used for the current study. S: Laser source; T1-T8: Transducers.
(b) Schematic of the coronal section/plane for the transducer scanning.
laser (LOTIS, Minsk, Belarus) with a repetition rate of 10 Hz and a pulse width
of <10 ns, illuminates the distal interphalangeal (DIP) joint of interest. The light
beam is guided via an optical fiber and the laser energy at the DIP joint is con-
trolled around 10 mJ/cm 2, which is far below the safety standard of 22 mJ/cm 2.
The ultrasound detection array (Figure 8.1a) is composed of eight 1 MHz trans-
ducers (Valpey Fisher, Hopkinton, MA) arranged equally spaced along a 210° arc
arm. The position and performance of each transducer in the ultrasound array was
calibrated carefully with controlled tissue phantom experiments. During an exam,
the palmar side of the DIP joint faced up, allowing the DIP joint to be illuminated
from the dorsal side of the finger. Our experience indicates that this way of light
illumination can give us maximized tissue penetration. To secure the position of
the finger, rubber bands were applied to the distal tip and the proximal end of the
finger. Both the ultrasound array and the examined finger were immersed in a
water tank for minimized ultrasound attenuation. The detected acoustic signals
were fed into preamplifiers and converted to digital signals by a multichannel A/D
board (PREAMP2-D and PCIAD1650, US Ultratek, Concord, CA). Labview pro-
gramming was used to control the entire examination procedure. With the ultra-
sound array, the photoacoustic exam took just about 30 s, allowing the collection of
acoustic signals at 40 detection positions (5 array positions) along a coronal plane
of the DIP joint (Figure 8.1b).
The qPAT reconstruction method detailed in Section 2.1 was used to quantita-
tively recover the optical absorption coefficient images of the DIP joints.
8.1.1.1.2 Patient Examination
Six subjects (two OA patients and four healthy controls) were enrolled in the study.
All participants (white females; mean age 61 years, range 45 to 71 years) provided
informed consent as part of the protocol approved by the Institutional Review Board
of University of Florida (UF). Participants were recruited from the Rheumatology
Clinic at the Shands Health Center of UF. One DIP finger joint, joint II of the left
hand, from each subject was examined clinically and photoacoustically. The DIP
joint is mostly vulnerable to OA disease and easily accessible with our current pho-
toacoustic scanning system.
Clinical examination of each patient was performed independently by a single
rheumatologist. Patients with OA were identified by clinical history and main clini-
cal features, including symptoms (predominantly pain and stiffness), functional
impairment and signs (joint enlargement and redness). The healthy controls had no
known OA or other joint diseases.
8.1.1.2 Results and Discussion

The in vivo 2D photoacoustic images were reconstructed using a dual mesh scheme (see
Section 3.1)—the fine mesh used for the forward calculation consisted of 5977 nodes
and 11,712 elements, while the coarse mesh used for the inverse calculation had 1525
nodes and 2928 elements. Based on the recovered 2D images, the optical absorption
coefficient of different joint tissues (bone, cartilage, and synovial fluid) and the struc-
tural size of joint cavity (cartilage and synovial fluid) were extracted. These parameters
are used as indicators/markers for differentiation of OA from normal joints in this study.
Figures 8.2a to 8.2f present the recovered absorption coefficient images (coronal
sections) for the six subjects examined. We can see that the bones (the red regions
5
5 5
0
0 0
–5 –5 –5
0.035
–5 0 5 –5 0 5 –5 0 5 0.03
(a) (b) (c) 0.025
0.02
5 0.015
5 6
0.01
4
0.005
2 0
0
0
–2
–5 –4 –5
–6
–5 0 5 –4 –2 0 2 4 –8 –6 –4 –2 0 2
(d) (e) (f )
FIGURE 8.2 Recovered absorption coefficient images for the six subjects examined. (a)–(d)
Normal joints. (e)–(f): OA joints. The color scale indicates the absorption coefficient (mm–1).
with highest absorption) are clearly delineated from the adjacent tissues (cartilage
and fluid) in the joint cavity for all the cases. While each joint has different shape
and size due to their positions relative to the scanning plan, the joint tissue/space
(indicated by the arrow) is readily identifiable. Compared with the healthy joints
(Figures 8.2a to 8.2d), we note that the value of absorption coefficient in the joint
cavity is elevated for the OA joints (Figures 8.2e and 8.2f). It is also notable that the
joint space narrowing appears to be evident for the OA joints.
The average absorption coefficient of joint tissues (cartilage, fluid, and bone) for
both the normal and OA joints were calculated and presented in Figures 8.3a and
8.3b. We can immediately tell from Figure 8.3a that both the diseased cartilage and
synovial fluid have significantly increased absorption coefficient compared with the
0.04
Cartilage
Fluid
0.03
Average µa (mm–1)
0.02
0.01
0.00
H1 H2 H3 H4 OA1 OA2
Case Number
(a)
1.0
µ(c)
a /µa
(b)
(f )
µa /µa(b)
Ratio of Absorption Coefficient
0.8
0.6
0.4
0.2
0.0
H1 H2 H3 H4 OA1 OA2
Case Number
(b)
FIGURE 8.3 (a) Average absorption coefficient of cartilage (light gray) and fluid (dark gray)
for the healthy (H1–H4) and osteoarthritic (OA1–OA2) joints. (b) Ratio of absorption coeffi-
cient of cartilage to that of bone (red), or ratio of absorption coefficient of fluid to that of bone
(blue) for the healthy and OA joints.
normal cartilage and fluid. While we note that the value of absorption coefficient of
bone for the OA joints is just slightly larger than that for the normal joints, the ratio
of the absorption coefficient of the cartilage or synovial fluid to that of the bone is
striking, as can be seen from Figure 8.3b. These quantitative optical changes asso-
ciated with the diseased joints are certainly attributed to the increased turbidity of
synovial fluid, collagen fibril network deterioration, and metabolic dysfunction of
bone, as elaborated previously.
8.1.2 2D Multispectral Quantitative PAT

The PAT system used is the same as described in Section 8.2.1. The only difference
here is that four laser wavelengths (720, 760, 805, and 850 nm) were used for multi-
spectral imaging. The qPAT method described in Section 2.5 was used to quantita-
tively recover the images of tissue functional parameters including oxy-hemoglobin
(HbO2), deoxy-hemoglobin (Hb), and water content (H2O) and acoustic velocity of
the DIP joints. The same two OA patients and four healthy controls described in
Section 8.2.1.1 participated this study.

The in vivo 2D photoacoustic images were reconstructed using a dual mesh scheme
(Section 3.1)—the fine mesh used for the forward calculation consisted of 23,665
nodes and 46,848 elements, while the coarse mesh used for the inverse calcula-
tion had 1525 nodes and 2928 elements. Based on the recovered 2D images, the
Hb, HbO2, oxygen saturation (= HbO2/(Hb+HbO2)), H2O and acoustic velocity of
phalanx bone and soft joint tissues in the joint space/cavity were extracted. These
parameters are used as indicators/markers for differentiation of OA from normal
joints in this study.
Figures 8.4 and 8.5 present the recovered sagittal images of Hb, HbO2, H2O and
acoustic velocity for a typical normal (Figures 8.4a, 8.4c, 8.5a, and 8.5c) and OA
(Figures 8.4b, 8.4d, 8.5b, and 8.5d) joints examined. Various joint structures are
identified in these images, in comparison with a sagittal MRI of a model normal DIP
joint (not photoacoustically scanned) shown in Figure 8.6. From Figures 8.4 and 8.5,
the structures shown in Hb, HbO2 and H2O are similar, where distal phalanx (dp),
middle phalanx (mp), joint cavity (jt), volar plate (vp), and extensor expansion (ee)
are all visualized for both the normal and OA joints. Skin and subcutaneous soft
tissues (su) is seen for the OA joint. Interestingly, the images of acoustic velocity
(Figures 8.5c and 8.5d) have very different appearance compared with the func-
tional/physiological images (Figures 8.4a to 8.4d and 8.5a to 8.5b)—the extensor
expansion is better visualized and flexor digitorum profoundus tendon (fdpt) is read-
ily identifiable while volar plate is not delineated from the velocity images. Nail and
subungual space (n and sus) is also seen in the velocity images. The finding that
the acoustic velocity is sensitive to tendon tissue is in fact consistent with that from
conventional ultrasound imaging of finger joints (Terslev et al. 2004). While the
photoacoustic images shown here have lower resolution relative to MRI, they allow
us to see most joint structures available from MRI of a DIP joint.
12
4.5
10 10
4 10
5 3.5 5
dp 8
3 dp
jt 2.5 ee jt vp
0 vp 0 su 6
ee mp 2 mp
–5 1.5 –5 4
1
2
–10 0.5 –10
–10 –5 0 5 10 –10 –5 0 5 10
(a) (b)
40
10 10 35
20
30
5 5
dp 15 dp 25
ee jt vp ee jt
0 0 vp 20
su
mp vp 10 mp
15
–5 –5
5 10
–10 –10 5
–10 –5 0 5 10 –10 –5 0 5 10
(c) (d)
FIGURE 8.4 Recovered Hb (a, b) and HbO2 (c, d) images for typical normal (a, c) and OA
(b, d) joints. The color scale indicates the Hb or HbO2 (μM). dp: distal phalanx; mp: middle
phalanx; jt: joint cavity; vp: volar plate; ee: extensor expansion; su: skin and subcutaneous
soft tissues.
Qualitatively, the structural differences between OA and normal joints are not
apparent from Figures 8.4 and 8.5; however, quantitative differences in functional
parameters between OA and normal joints are striking from the extracted aver-
age oxygen saturation and water content of the joint and phalanx bones given in
Figures 8.7 and 8.8. We can immediately tell from Figure 8.7 that both the diseased
joint and bones have significantly decreased oxygen saturation compared with the
normal ones. In particular, the oxygen saturation of the OA joints is smaller than
72%, while it is larger than 79% for the normal joints. Figure 8.8 shows that the
OA joints have elevated water content especially for the joint cavity. The average
acoustic velocity of the joint and bones is given in Figure 8.9, where we can see that
the acoustic velocity of the joint tissues is notably larger for the OA joints than that
for the normal joints, while the value of this quantity for bones is indistinguishable
among the OA and normal joints.
Changes in cartilage, underlying subchondral bone, and synovial fluid are the
tissues most affected by OA (these three types of tissues are the major components
60
10 40 10
50
35
5 5
d 30 40
ee vp
d
jt 25 ee
0 0 jt
30
m vp 20
vp su
m
–5 15 –5 20
10
–10 10
–10 5
–10 –5 0 5 10 –10 –5 0 5 10
(a) (b)
× 106 × 106
2
10 10 2
n & sus 1.9
n & sus 1.9
5 dp 1.8 5
ee 1.8
ee dp
jt 1.7
0 0 jt 1.7
mp tdpt mp
1.6 1.6
–5 su –5 tdpt
1.5 su 1.5
–10 1.4 –10 1.4
–10 –5 0 5 10 –10 –5 0 5 10
(c) (d)
FIGURE 8.5 Recovered H2O (a, b) and acoustic velocy (c, d) images for a typpical normal (a, c)
and OA (b, d) joints. The color scale indicates the H2O (%) or acoustic velocity (m/s). dp: distal
phalanx; mp: middle phalanx; jt: joint cavity; ee: extensor expansion; su: skin and subcutaneous
soft tissues; fdpt: flexor digitorum profoundus tendon; n and sus: nail and subungual space.
n
sus
sus
dp fd pt dp
vp fdpt
vp
ee
fdpt
FIGURE 8.6 Sagittal MRI of a model normal DIP joint. dp: distal phalanx; mp: middle
phalanx; jt: joint cavity; vp: volar plate; ee: extensor expansion; su: skin & subcutaneous
soft tissues; fdpt: flexor digitorum profoundus tendon; n and sus: nail and subungual space.
85
Joint
Phalanx bones
H: Healthy Subject
OA: Osteoarthritic Subject
80
Oxygen Saturation (%)
75
70
65
H1 H2 H3 OA1 OA2
Case Number
FIGURE 8.7 Average oxygen saturation of the joint (light gray) and phalanx bones (dark
gray) for the healthy (H1–H4) and osteoarthritic (OA1–OA2) joints.
80
Joint
Phalanx bones
70
Water Content (%)
60
50
40
30
H1 H2 H3 OA1 OA2
Case Number
FIGURE 8.8 Average water content of the joint (light gray) and phalanx bones (dark gray)
for the healthy (H1–H4) and osteoarthritic (OA1–OA2) joints.
2.4
Joint
2.2 Phalanx bones
Velocity (×106 mms–1)
2.0
1.8
1.6
1.4
H1 H2 H3 OA1 OA2
Case Number
FIGURE 8.9 Average acoustic velocity of the joint (light gray) and phalanx bones (dark
gray) for the healthy (H1–H4) and osteoarthritic (OA1–OA2) joints.
of the joint or joint cavity defined in this paper) (Gerwin et al. 2006, Blanco et al.
2004). Bone is frequently remodeled in OA, with new bone formed either as osteo-
phytes at the margins of the articular cartilageor as a change in subchondral bone
density (Gerwin et al. 2006, Blanco et al. 2004). These changes in bone structure can
allow blood flow to more easily penetrate the subchodral bone plate and the calcified
cartilage that lies between subchondral bone and cartilage (Gerwin et al. 2006, Blanco
et al. 2004). There is now cumulative scientific evidence suggesting that the trigger
for many of these changes is the low oxygen levels in the diseased tissues includ-
ing the cartilage of an OA joint (Blanco et al. 2004; Henrotin et al. 2005). These
low oxygen levels will cause the formation of new blood vessels, allowing the dis-
eased synovium to invade the surrounding tissues (Blanco et al. 2004). It seems these
changes correlate well with the decreased oxygen saturation seen in the joint cavity
of OA joints in our study. The mild inflammatory changes frequently seen in the
synovial fluid of patients with OA would also produce high water content in the joint
cavity, a feature that is also seen in our measurements (Figure 8.8). In addition, nor-
mal cartilage is tough, elastic, and very durable and comprises mainly collagen (20%
to 30%) and water molecules (67% to 80%). Early changes in OA cartilage include
proteoglycan loss, with a resultant increase in hydration of the articular cartilage.
These changes also likely contribute to the increase in water content and decrease in
oxygen saturation seen in the joint cavity of OA joints. While the changes in water
content and hemoglobin concentration in the joint cavity may all contribute to the
change in acoustic velocity associated with an OA joint, a full understanding on the
mechanism behind this change warrants future studies.
8.1.3 3D Single-Spectral Quantitative PAT

8.1.3.1.1 3D Quantitative Photoacoustic Imaging
The PAT system used is the same as that described in Section 8.2.1. The acoustic
signals at 360 scanning locations over a spherical detecting surface surrounding the
DIP joint of interest were collected through the combined rotations of two rotary
stages with orthogonal rotary axes. The scanning procedures can be described as
follows. One rotary stage controls the ultrasound detecting array to scan 40 circular
locations with an interval of 6° in the plane of ultrasound array, covering a total of
240° along the circle surrounding the distal end of the finger joints, while the other
rotary stage tilts the ultrasound detecting array from –60° to 60° with the horizontal
plane by a step of 15°, allowing for nine detecting angles and a 240° coverage along
the cross-section of the finger joint. With the ultrasound array, the whole examina-
tion takes 5 minutes.
The optical absorption coefficient images of the DIP joints were quantitatively
reconstructed using a finite element mesh of 7519 nodes and 41,323 tetrahedral ele-
ments with a 3D qPAT reconstruction algorithm similar to the 2D qPAT method
described in Section 2.1. In brief, the photon density can be calculated from the 3D
photon diffusion equation and the estimated optical absorption coefficients are itera-
tively updated by dividing the absorbed optical energy reconstructed in conventional
PAT by the calculated photon density, until the product of the calculated photon den-
sity and the current estimated optical absorption coefficients is close enough to the
absorbed optical energy reconstructed. A full-width at 30% maximum (FW30%M)
method adapted from optical imaging techniques was used to quantitatively analyze
the absorption coefficient profiles from the recovered 3D images, which is capable
of improving the measurement accuracy in determining the thickness of optically
imaged joint tissues (Jiang 2010). In brief, transect lines that are parallel to the mid-
dle line of the bones and that are distributed along the entire bone area with an
interval of 0.5 mm were chosen to interpolate the optical absorption profile for each
sectional slice. The interpolated absorption curve for each transect line was divided
into three different joint tissues (bone, cartilage, and fluid) and the interface points
among adjacent joint tissues were taken as 30% of the nearest maximum values
along the curve. By using this FW30%M threshold, the mean thickness of the joint
and the averaged optical absorption coefficients of joint tissues (bone, cartilage, and
fluid) can be calculated in each sectional slice. The volumetric mean values of the
joint thickness and the optical absorption coefficients of joint tissues can be further
obtained by averaging the corresponding values in all sectional slices. The volu-
metric mean absorption coefficient of different joint tissues, the structural size, and
the lowest sectional absorption coefficient of the joint cavity from both OA joints
and healthy controls are used as indicators/markers for differentiation of OA from
normal joints in this study. These indicators/markers have been previously employed
by radiography or pure optical imaging methods to detect osteoarthritis/joint rheu-
matism (Peloschek et al. 2007, Jacobson et al. 2008, Yuan et al. 2008, Jiang 2010).
8.1.3.1.2 Patient Examination
The same six subjects described in Section 8.2.1 were used for this 3D PAT study.

Consecutive coronal and sagittal slices from the recovered 3D images for a typical
normal joint are presented below. The exact locations of these selected slices may be
slightly different from subject to subject. The coronal and sagittal slices for a normal
joint are shown in Figures 8.10a and 8.10b. As can be observed from Figures 8.10a
and 8.10b, the bones are clearly delineated from the adjacent tissues (cartilage and
fluid) in the joint cavity, which is clearly identified in both the coronal and sagittal
sections. Further observation finds that the synovial fluid seems to be differentiable
from the surrounding cartilages in some slices (Z = 7.5 mm, Z = 8 mm in the coro-
nal section and almost all the slices in the sagittal section). Figures 8.11a and 8.11b
display selected coronal and sagittal slices from the recovered 3D images for an
OA joint. Again, the bones are clearly differentiable from the adjacent joint tissues
(cartilage and fluid), and the joint cavity is easily identified in both the coronal and
sagittal sections.
The average absorption coefficient and structural size of joint tissues for all the
subjects are calculated and presented in Table 8.1. We note from Table 8.1 that the
recovered volumetric mean absorption coefficient of the diseased synovial fluid
ranges from 0.022 to 0.023 mm–1, while it is below 0.017 mm–1 (0.008 to 0.017 mm–1)
for the normal fluid. Difference between OA joints and healthy controls is also
apparent, in terms of the volumetric mean absorption coefficient of the synovial fluid
in the joint cavity. The difference is more striking when the ratio of the volumetric
mean absorption coefficient of the synovial fluid to that of the bone is considered
(0.58 to 0.63 for OA joints vs. 0.25 to 0.46 for normal joints). For the diseased carti-
lage, the volumetric mean absorption coefficient (0.026 to 0.028 mm–1) is generally
larger than that of normal joints (0.015 to 0.024 mm–1). The ratio of the volumetric
mean absorption coefficient of the cartilage to that of the bone further confirms the
difference between OA and normal joints (0.72 to 0.76 for OA joints vs. 0.43 to 0.61
for normal joints).
Further observation from Table 8.1 indicates that the lowest sectional absorption
coefficient in the joint cavity ranges from 0.014 to 0.018 mm–1 in the diseased joints,
while it is in the range of 0.004 to 0.010 mm–1 for the normal joint. A more significant
difference between the OA and normal joints is observed in the ratio of the lowest
sectional absorption coefficient of the joint cavity to the volumetric mean absorption
coefficient of the bone (0.38 to 0.50 for OA joints vs. 0.11 to 0.26 for normal joints).
The difference was found to be as large as 2:1 in the OA joints compared to the normal
joints.
We also note from Table 8.1 that the measured thickness of the joint cavity (synovial
fluid and cartilage) for OA joints ranges from 1.1 to 1.5 mm, with a mean value of 1.3 mm,
whereas the range of this size for healthy joints is from 1.7 to 2.5 mm, with a mean value of
2.0 mm. Compared with the normal joints, the dimensional narrowing of the joint cavity
is obvious for OA joints, which is in agreement with typical radiographic evidences of OA
disease (Peloschek et al. 2007, Jacobson et al. 2008).
Z = 5.5 mm Z = 6 mm Z = 6.5 mm Z = 7 mm Z = 7.5 mm Z = 8 mm
10
CL CL CL CL
DP DP DP DP DP DP
SF
SF 0.07
5
DIP DIP DIP DIP DIP DIP

0 0.06
Y (mm)
CL
–5 CL
0.05
IP IP IP
IP IP IP
–10
–10 –5 0 –10 –5 0 –10 –5 0 –10 –5 0 –10 –5 0 –10 –5 0 0.04
X (mm) X (mm) X (mm) X (mm) X (mm) X (mm)
(a) Coronal Section (dorsal palm)
0.03
X = –5.5 mm X = –5 mm X = –4.5 mm X = –4 mm X = –3.5 mm X = –3 mm
Clinical Applications and Animal Studies
10
DP DP DP DP DP DP 0.02
SF SF SF
SF SF SF
5
0.01
DIP DIP DIP
0
Y (mm)
CL DIP DIP CL CL
CL CL DIP CL 0
–5
IP IP IP IP IP
IP
0 5 10 0 5 10 0 5 10 0 5 10 0 5 10 0 5 10
Z (mm) Z (mm) Z (mm) Z (mm) Z (mm) Z (mm)
(b) Sagittal section (Medial Lateral)
FIGURE 8.10 Selected coronal (a) and sagittal (b) planes of the recovered 3D absorption coefficient image for a typical normal joint. DP: distal pha-
235
lanx; IP: intermediate phalanx; DIP: distal interphalangeal joint; CL: cartilage; SF: synovial fluid.
Z = –9 mm Z = –8.5 mm Z = –8 mm Z = –7.5 mm Z = –7 mm Z = –6.5 mm
10
236
DP DP DP CL DP DP 0.07
DP
SF
SF
5
DIP 0.06
Y (mm)
DIP DIP
0 DIP DIP CL
DIP
0.05
–5 IP
IP IP IP IP IP
–5 0 5 –5 0 5 –5 0 5 –5 0 5 –5 0 5 –5 0 5 0.04
X (mm) X (mm) X (mm) X (mm) X (mm) X (mm)
(a) Coronal Section (dorsal palm)

0.03
X = –2.5 mm X = –2 mm X = –1.5 mm X = –1 mm X = –0.5 mm X = 0 mm

0.02
10 DIP DIP DP DIP
DP DP DP DP DP
DIP
TE TE 0.01
5 TE
DIP DIP
Y (mm)
0 0
TE
–5
IP IP IP IP IP IP
–10 –5 0 –10 –5 0 –10 –5 0 –10 –5 0 –10 –5 0 –10 –5 0
Z (mm) Z (mm) Z (mm) Z (mm) Z (mm) Z (mm)
(b) Sagittal section (Medial Lateral)
FIGURE 8.11 Selected coronal (a) and sagittal (b) planes of the recovered 3D absorption coefficient image for a typical OA joint. DP: distal phalanx;
Photoacoustic Tomography
IP: intermediate phalanx; DIP: distal interphalangeal joint; CL: cartilage; SF: synovial fluid; TE: tendon.
TABLE 8.1
Averaged Absorption Coefficient and Size of Joint Tissues for 2 OA and 4
Normal Joints
Joint
Joint Mean Thickness µ (ac ) µ (af ) µa
µ (mm )
a
a
–1
µ (ab ) µ (ab ) µ (ab )
No. Tissues μa(mm–1) (mm)
Healthy 1 Bone 0.035 0.004 1.7 0.43 0.25 0.11
cartilage 0.015
fluid 0.008
2 Bone 0.039 0.010 2.1 0.61 0.45 0.26
cartilage 0.024
fluid 0.017
3 Bone 0.034 0.009 2.5 0.59 0.45 0.26
cartilage 0.020
fluid 0.0 IS
4 Bone 0.037 0.008 1.8 0.61 0.46 0.22
cartilage 0.023
fluid 0.017
5 Bone 0.036 0.018 1.5 0.76 0.63 0.50
cartilage 0.028
fluid 0.023
6 Bone 0.037 0.014 1.1 0.72 0.58 0.38
cartilage 0.026
fluid 0.021
a
a is the lowest sectional optical absorption coefficient in the joint cavity.
8.1.4 3D Multispectral Quantitative PAT

The PAT system used is the same as described in Section 8.2.1, and four laser
wavelengths (720, 760, 805, and 850 nm) were used for multispectral imaging. The
qPAT method described in Section 2.5 was adapted to quantitatively recover the 3D
images of tissue functional parameters including oxy-hemoglobin (HbO2), deoxy-
hemoglobin (Hb), total hemoglobin (Total-Hb), and oxygen saturation (SO2) of the
DIP joints. Four of the six recruited subjects (two OA patients and two healthy con-
trols) described previously completed the whole procedure of 3D multispectral pho-
toacoustic scanning and resulted in successful multispectral image reconstruction.

The 3D photoacoustic images of the DIP joints from all four human subjects were
reconstructed and quantitatively analyzed using 3D qPAT reconstruction method
with a spherical mesh of 7519 finite element nodes and 41,323 tetrahedral elements.
Figures 8.12a and 8.12b are the recovered optical absorption maps in coronal section
and sagittal section, respectively, of an OA finger joint at wavelengths from 730 to
880 nm. In Figures 8.12a and 8.12b, the joint tissue/space is differentiated from the
adjacent bones in the coronal and/or sagittal sections at several optical wavelengths
(760, 805, 850, and 880 nm). Relative structural consistency of the recovered optical
absorption images are observed between the wavelengths of 760, 805, and 850 nm,
which are selected to further quantitatively resolve the tissue chromophore concen-
trations. The reason that the joint tissue/space is missing or fused with the adjacent
bones at some optical wavelengths (such as 730 and 825 nm) is most likely because of
the motion error during the 30-minute examination procedure, despite an optimized
wavelength order is followed in our 3D multispectral PAT scanning. We believe that
the motion error and the structural inconsistency of the recovered optical absorption
maps at some wavelengths can be significantly reduced by using an ultrasound array
with larger number of ultrasound detectors (e.g., see the 192-element array system
described in Chapter 4).
The quantitatively resolved concentrations of HbO2 and Hb of the same finger
joint are displayed in Figure 8.13. The concentrations of HbO2 (Figure 8.13a) in dif-
ferent joint tissues are clearly differentiated in high resolution both in coronal section
and sagittal section, and it is relatively higher in the bones than in the joint cavity.
Figure 8.13b shows the concentrations of Hb both in coronal section and sagittal
section, where the joint cavity is differentiated from the adjacent phalanges in the
sagittal section and is barely identifiable in the coronal section. We believe that the
relatively poor quality of the resolved Hb concentration images in coronal section
is most likely due to (1) the low-quality image at the wavelength 760 nm, where the
absorption spectrum of Hb has a high peak, and (2) the potential motion error during
the 3D multispectral PAT scanning.
Figure 8.14 shows the quantitatively resolved concentrations of HbO2 and Hb
from a typical healthy finger joint. Again, the structure of joint space is clearly iden-
tified in the images of each resolved chromophore, and the concentrations of HbO2
and Hb in the joint cavity are found to be relatively lower than that in the adjacent
phalanges. In comparison with the OA finger joint shown in Figure 8.13, the joint
cavity in healthy subjects seems more distinctive.
For all four examined in vivo finger joints, the average concentrations of oxy-
hemoglobin (HbO2), total hemoglobin (Total-Hb), and the oxygen saturation (SO2)
in the joint cavity and the adjacent bones are calculated and presented in Figure 8.15.
Figure 8.15a shows that the average concentrations of HbO2 and Total-Hb in the
finger phalanges are 44~58 μM, and 68~96 μM, respectively. The quantitatively
resolved concentration levels of HbO2 and Total-Hb are in agreement with the
measurements by diffuse optical tomography (Jiang 2010). Further observations
indicate a difference in the levels of total-Hb between finger phalanges with OA
disease and healthy phalanges, which are 86~96 μM and 68~69 μM, respectively.
The finding of enhanced hemoglobin levels in OA joints has the pathologic roots
revealed by previous studies, where angiogenesis has been observed in diseased
730 nm 760 nm 805 nm

0.06 0.07
10 10 10
0.07 0.06
0.05
0.06
0.04 5 0.05
Y (mm)
Y (mm)
Y (mm)
5 0.05 5
0.04
0.04 0.03
0 0.03 0 0 0.03
0.02
0.02 0.02
–5 0.01 –5 0.01 –5 0.01
0 0
–5 0 5 –5 0 5 –5 0 5
X (mm) X (mm) X (mm)
825 nm 850 nm 880 nm
10 10 10 0.06
0.04 0.05
0.05
5 5 0.04 5
0.03
Y (mm)
Y (mm)
Y (mm)
0.04
0.03
0.03
0 0.02 0 0
0.02 0.02
0.01
–5 –5 0.01 –5 0.01
0 0
–5 0 5 –5 0 5 –5 0 5
X (mm) X (mm) X (mm)
(a)
730 nm 760 nm 805 nm
10 10 10 0.05
0.06
0.04
0.05 0.04
5 5 5
Y (mm)
Y (mm)
Y (mm)
0.04 0.03
0.03
0.03 0.02
0 0 0 0.02
0.02
0.01 0.01
–5 0.01 –5 –5
0 0 0
–10 –5 0 –10 –5 0 –10 –5 0
Z (mm) Z (mm) Z (mm)
825 nm 850 nm 880 nm
10 0.035 10 10 0.035
0.04
0.03 0.03
5 0.025 5 0.03 5 0.025
Y (mm)
Y (mm)
Y (mm)
0.02 0.02
0 0.015 0 0.02 0 0.015
0.01 0.01
0.01
–5 0.005 –5 –5 0.005
–10 –5 0 –10 –5 0 –10 –5 0

Z (mm) Z (mm) Z (mm)
(b)
FIGURE 8.12 Recovered optical absorption images in coronal section (a) at z = –8 mm and
sagittal section (b) at x = –1 mm of an in vivo finger joint for light source in six wavelengths.
Z= –8 mm X= –1 mm
10 10 70
80 60
5 5 50
60
Y (mm)
Y (mm)
40
0 40 0 30
20
20
–5 –5 10
0 0
–5 0 5 –10 –5 0
X (mm) Z (mm)
(a)
10 10
50 50
40 40
5 5
Y (mm)
Y (mm)
30 30
0 0
20 20
10 10
–5 –5
0 0
–5 0 5 –10 –5 0
X (mm) Z (mm)
(b)
FIGURE 8.13 Concentrations of oxy-hemoglobin (a) and deoxy-hemoglobin (b) in μM at

coronal section (z = –8 mm) and sagittal section (x = –1 mm) of examined finger joint from
an OA patient.
tissues. The oxygen saturations of the four in vivo finger joints are observed from
58 to 66% in the finger phalanges, while finger phalanges with OA disease seems to
have lower oxygen saturations levels (55~60%) than healthy phalanges (65~66%).
Hypoxia in osteoarthritic phalanges has been revealed and seemed to appear in
prearthritic stage by previous studies of finger joints diseases (Jeon et al. 2008).
Our results are in good agreement with the previous findings of hypoxia in osteo-
arthritic phalanges.
As shown in Figure 8.15b, the average concentrations of HbO2 and Total-Hb
in the finger joint cavity are 16~29 μM, and 27~53 μM, respectively. An obvious
Z = 2 mm X = –0.5 mm
60
10 10 60
50
50
5 40 5
40
Y (mm)
Y (mm)
30 30
0 0
20 20
–5 10 –5 10
0 0
–5 0 5 –5 0 5
X (mm) Z (mm)
(a)
35
35
10 10
30
30
25
5 25
5
20
Y (mm)
Y (mm)
20
15 0 15
0
10 10
–5 5 –5 5
0 0
–5 0 5 –5 0 5
X (mm) Z (mm)
(b)
FIGURE 8.14 Concentrations of oxy-hemoglobin (a) and deoxy-hemoglobin (b) in μM at

coronal section (z = 2 mm) and sagittal section (x = –0.5 mm) of examined finger joint from
a healthy subject.
difference in the level of total-Hb between joint cavities with OA disease and
healthy joint cavities can be observed in Figure 8.15b, which are 46~53 μM and
27~42 μM, respectively. The oxygen saturations in the joint cavities are observed
to be 55~69%, and osteoarthritic fingers seem to have lower oxygen saturations
levels (55~59%) than healthy phalanges (59~69%) in the joint cavities. Again, the
enhanced hemoglobin levels and dropped oxygen saturations in OA joint cavities
are in agreement with the pathologic angiogenesis and hypoxia related with finger
joint diseases.
100 100
Concentrations of HbO2 and Total-Hb (μm)

HbO2
90 Total-Hb 90
SO2
80 80
Oxygen Saturation SO2 (%)

70 70
60 60
50 50
40 40
30 30
20 20
10 10
0
1 2 3 4
Subject No.
(a)
100 100
Concentrations of HbO2 and Total-Hb (μm)
HbO2
90 Total-Hb 90
SO2
80 80
Oxygen Saturation SO2 (%)

70 70
60 60
50 50
40 40
30 30
20 20
10 10
0
1 2 3 4
Subject No.
(b)
FIGURE 8.15 Average concentrations of oxy-hemoglobin (HbO2), total hemoglobin

(Total-Hb) and the oxygen saturation (SO2) in finger joint bones (a) and in joint cavity (b).
8.2 INTRAOPERATIVE IMAGING
Breast cancer remains a major public health problem affecting one in four women in
the United States (Jemal et al. 2008). As a result of population-based screening pro-
grams for breast cancer, most women are currently diagnosed with early stage breast
cancer. In a lumpectomy, margin is typically examined using pathological methods
to determine if there is at least 2 mm of surrounding normal breast tissue beyond
the tumor border; this is considered to be a negative margin. However, if the tumor
cells come to the edge of the resected tissue, it is considered to be a positive margin
(Taghian et al. 2005, Assersohn et al. 1999).
Methods currently available for intraoperative margin assessment, such as gross

examination, frozen section, ultrasound, and touch-prep, have various limitations
with false-negative diagnoses in 20% to 50% of the patients (Balch et al. 2005,
Sigal-Zafrani et al. 2004, Meric et al. 2003). Therefore, there is an urgent need to
develop sensitive and accurate intraoperative methods for detecting tumor margins
in order to reduce tumor recurrence and to improve the survival rate of breast
cancer patients.
Here we evaluate high-resolution imaging of tumor margins and resection inspec-
tion by using an intraoperative photoacoutic imaging system. The penetration depth
for our intraoperative photoacoutic imaging system is 2.3 mm, which is clinically
relevant for tumor margin assessment of lumpectomy specimen. We demonstrate
this technique using a tumor-bearing animal model. The results obtained show accu-
rate correlation with the histology.
8.2.1 Intraoperative Photoacoutc Tomography (iPAT) System

In our microelectromechanical system (MEMS)-based photoacoustic imaging
system, short laser pulses of 6 ns duration are generated from a Nd:YAG 532 nm
laser (NL 303HT from EKSPLA, Lithuania), and a reflection mirror is employed
to deliver light into a miniaturized probe (Figure 8.16a). Inside the probe, there are
an aperture (for generating a 0.2 mm-diameter light spot), a wedge-shaped mir-
ror (Figure 8.16c) and a MEMS mirror (Figure 8.16b). A custom-designed PZT
(DL-53HD, DeL Piezo Specialties, LLC, Wellington, FL) ultrasound transducer
(unfocused ring-shape, 5.5 MHz central frequency, 7 mm inner diameter and
11.5 mm outer diameter) is attached to the end of the probe to detect photoacoustic
waves.
A MEMS mirror based on electrothermal bimorph actuation is utilized to real-
ize 2D light scanning (Wu and Xie 2008). The movable mirror plate (Figure 8.16b)
is as large as 0.8 × 0.8 mm in a 2 × 2 mm device footprint. The MEMS mirror is
fixed on a ramp with an angle of 22.5° (Figure 8.16d), allowing light to scan from
position 3 to position 4 (Figure 8.16c), while the MEMS mirror moves from posi-
tion 1 to position 2. The measured vertical resonant frequency is 500 Hz. Four
channels of voltage signals are utilized to drive four actuators (Figure 8.16b) of the
MEMS mirror.
The capability of our iPAT system was validated by mouse experiments involv-
ing tumor extirpation (n = 8). Figure 8.18a shows the photograph of breast tumor in
a mouse. The laser beam illuminated the tumor surface along the z-axis at an energy
density of 8 mJ/cm2, which is lower than the American National Standards Institute
safety limit of 20 mJ/cm2. Two 0~4 V ramp signals with repetition frequency of
0.1 Hz were used to drive two actuators moving along the y axis and the other two
signals with the same magnitude but different repetition frequency of 0.002 Hz were
used to drive the actuators moving along the x-axis. The photoacoustic signals were
recorded without averaging at 10 Hz frame rates, amplified and converted to a 1D
image assuming a constant ultrasound velocity in soft tissue (1.54 mm/μs). A square
area (10 × 10 mm) in x–y plane was imaged to produce a 3D image of the tissue
volume.
Aperture Footprint
Fixed Mirror
Actuator
Transducer
Reflection
Mirror
MEMS Mirror
Plastic Film Electrodes
(a) (b)
Light
22.5°
Fixed Mirror
Transducer 2
MEMS Mirror
3 4
Tissue 1
(c) (d)
FIGURE 8.16 Schematic representation of the MEMS probe. (a) 3D rendering of the probe.
(b) Photograph of the MEMS mirror with four actuators and five pads bonded with gold wire.
(c) Schematic of the miniaturized probe. A wedge-shaped mirror plated with aluminum is
fixed in position to redirect the light beam to the MEMS mirror. (d) The MEMS mirror is
fixed on a ramp with an angle of 22.5°.
One set of data measured from a cylindrical phantom with ink covering the
imaged surface was used to calibrate all the experimental data used in this study
to reduce the effect of focal signal from the center area from the ring-shaped trans-
ducer. All slices were processed and displayed by normalizing the images to the
same scale (0–256) with a threshold (–6dB level of the normalized peak photo-
acoustic signal).

A 100 µm-diameter tissue mimicking target (absorption coefficient μa = 0.049 mm–1,
reduced scattering coefficient μ′s.= 0.5 mm –1) embedded in turbid media (μa.=
0.012 mm–1, μ′s.= 0.35 mm–1) was imaged. Chicken breasts with thickness of up to 2.3
mm were added to investigate the resolution and signal-to-noise ratio (SNR) versus
1
Simulation (Transverse)
0.9
Transverse
0.8 Axial
Resolution (mm)
0.7
0.6
0.5
0.4
0.3
0.2
0 0.5 1 1.5 2 2.5
Distance (mm)
(a)
SNR
40
30
SNR (dB)
20
10
0
0 0.5 1 1.5 2 2.5
Distance (mm)
(b)
FIGURE 8.17 Performance of this system. (a) Axial and transverse resolutions versus target
depth. (b) Signal-to-noise ratio (SNR) versus target depth.
target position. In Figure 8.17a, the system’s axial and transverse resolutions are plot-
ted versus the target’s position. The experimental axial resolution (black crosses),
determined by the central frequency of transducer, varied from 0.36 to 0.4 mm, which
agrees well with the theoretical value of 0.3 mm. The experimental transverse reso-
lution, based on the size of light spot, decreased from 0.24 to 0.96 mm when the
depth increased. Again this is consistent with the theoretical depth-dependent trans-
verse resolution of 0.21~0.95 mm. The SNR measured in turbid media is shown in
Figure 8.17b, decreasing from 32 to 18 dB within 2.3 mm.
The volumetric shape, size, and position of the tumor are shown in Figure 8.18b.
The surgeon removed the tumor marked with T in Figure 8.18b. In these experi-
ments, the line of surgical resection of tumor was intentionally carried out beyond the
margin of the detectable tumor, consistent with clinical practice. After the surgical
y (c) y
(a)
x x
(e)
y (d) y
(b)
x z x z
T
10 mm 10 mm
10 mm 7 mm
10 mm 7 mm
FIGURE 8.18 In vivo 3D tumor mapping in a mouse model. (a) Photograph of the mouse
with tumor implanted in abdomen before surgery. (b) In vivo 3D photoacoustic image. The
distance shown along z-axis includes the tumor depth and the distance between the surface
of transducer and mouse skin. (c) Photograph of the mouse after tumor resection. (d) 3D
photoacoustic image of the tumor cavity after surgery to examine the completeness of the
tumor resection. (e) Photograph of the tumor resection guided by 3D photoacoustic image
shown in (b).
procedure, the surgical area shown in Figure 8.18c was inspected again by the same
imaging method to ensure the complete resection of tumor. The image shown in
Figure 8.18d confirmed the completed resection of tumor. The signal (indicated by
the arrow in Figure 8.18d) was from the inner surface of the ring-shaped transducer.
This signal disappeared in Figure 8.18b due to strong signals from the tumor, and
normalized image procedure was used.
Following tumor removal, the tumor size as determined by iPAT was compared
with the actual tumor size as measured histologically. Hematoxylin and eosin (H&E)
stained sections along the x-z plane were obtained after the photoacoustic experi-
ments. A 2D PAT slice along the red dashed line in Figure 8.18b was selected and
compared with the corresponding H&E stained section (the black dashed line in
Figure 8.18a) as shown in Figures 8.19a and 8.19b. Figure 8.19c shows the photo-
graph of the tumor in comparison with the H&E and PAT slices. Also, the red dashed
line along the top margin of the tumor shown in Figure 8.19a matches well with the
top margin (black dashed line in Figure 8.19b) of the H&E section and photograph
(red dashed line in Figure 8.19c). The size along two typical directions (Lines 1 and
2 in Figure 8.19a) was estimated to be 2.1 and 1.2 mm, respectively, in excellent
agreement with the actual dimensions of 2.2 and 1.1 mm measured from the histol-
ogy slice (Lines 1′ and 2′ in Figure 8.19b). Such comparisons were applied to other
two H&E staining sections along the x-z plane for this mouse, and we found that the
measured error was less than 8.5% for all three slices.
(a) (b) (c)
1'
1 2'
2 x
z
(d) (e) (f)
1'
1 2'
2 x
y
FIGURE 8.19 Quantitative analysis of the photoacoustic slices and H&E stained sections.
(a) Photoacoustic slice (along the red dashed line in Figure 8.18b). Lines 1 and line 2 indicate
the imaged dimensions along two different directions. (b) H&E stained section in the same
position. Lines 1′ and line 2′ denote the measured dimensions in the H&E stained section.
(c) Photograph of the tumor. All the lines indicated by arrows in (a), (b), and (c) show the top
margins of the tumor. (d) Transverse photoacoustic slice from another mouse. Lines 1 and
line 2 indicate the imaged dimensions along two different directions. (e) H&E stained section
in the same position. Lines 1′ and line 2′ present the actual dimensions in the H&E stained
section. (f) Photograph of the tumor. All the lines indicated by arrows in (d), (e), and (f) pres-
ent the margins of the tumor in transverse plane. Scale bar, 500µm.
In another independent mouse tumor experiment, photoacoustic imaging demon-

strated tumor nodules as shown in Figure 8.19d. The tumor size imaged by PAT along
lines 1 and 2 (Figure 8.19d) was 2.5 and 1.9 mm, showing an accurate measurement
of the actual size of 2.5 and 2.1 mm, measured by histology along lines 1′ and 2′
(Figure 8.19e). In this case, the largest observed error for the remaining two H&E
sections was 9.5%.
Photoacoustic slices with histological correlations from the third mouse are shown
in Figures 8.20a and 8.20b. The structures indicated by the arrow in Figure 8.20b
are consistent with each other. After quantitative analysis, the largest observed error
was 6.5%. Figures 8.20c and 8.20d show the PAT slices and histological correla-
tions for the fourth mouse where we added 1.1-mm-thick chicken breast on top of
the tumor to simulate the real specimen resected from the human breast. The 3D top
margins of the tumor were clearly seen in the PAT images and consistent with the
H&E sections. The white dashed line shows the surface of the chicken breast. The
error between the PAT image and H&E sections was found to be less than 16.5%.
A 1.8-mm-thick chicken breast was added to the top of the fifth mouse and the PAT
images with the associated H&E section are shown in Figures 8.20e and 8.20f. The
tumor shape imaged by PAT was close to that from the H&E sections, but the error
increased to 21.5% in this case.
(a) (b) (c)
(d) (e) (f )
FIGURE 8.20 Quantitative analysis of photoacoustic images in correlation with H&E sec-
tions with increasing image depth. (a, b) PAT slices and H&E sections for a tumor beneath
the skin of mouse (0.3 mm depth). (c, d) PAT slices and H&E sections for a tumor beneath
the mouse skin and a 1 mm-thick chicken breast. (e, f) PAT slices and H&E sections for a
tumor beneath the mouse skin and a 1.8-mm-thick chicken breast. White dashed lines show
the surface of chicken breast. Scale bar, 500 μm.
8.3 BRAIN IMAGING
Epilepsy is a common, chronic, neurological disorder characterized by seizures.
Three percent of people will be diagnosed with epilepsy at some time in their lives
(Hauser et al. 1996). Indeed, approximately 50 million people worldwide have
epilepsy, and 20% to 30% of these patients are refractory to all forms of medical
treatment (Hauser 1992). Seizures are transient signs and symptoms of abnormal,
excessive, or hypersynchronous neuronal activity in the brain (Fisher et al. 2005). In
most cases, seizures are controlled, although not cured, with anticonvulsant medi-
cation. Seizure types are classified according to whether the source of the seizure
is localized (partial or focal onset seizures) or distributed (generalized seizures)
(Greenfield et al. 2011). Localization-related epilepsies arise from an epileptic focus.
A partial seizure may spread within the brain, a process known as secondary gen-
eralization. Generalized epilepsies, in contrast, arise from many independent foci
(multifocal epilepsies) or from epileptic circuits that involve the whole brain. In epi-
lepsies of unknown localization, it remains unclear as to whether they arise from a
portion of the brain or from more widespread circuits.
For those patients with medically intractable focal epilepsy, the best treatment
option is resective brain surgery (Berg et al. 2007). Although several factors can
impact the success of epilepsy surgery, the primary reason of failure is the incomplete
mapping of the local epilepsy network, which results in incomplete resection of epi-
leptogenic foci (Engel 2004). Much of the attention on epilepsy surgery has been
directed at identifying single neuronal populations. This approach has, in many
cases, led to failed surgical outcomes, because seizures typically involve groups of
neurons interacting both locally and across several cortical and subcortical brain
regions. A better understanding of the regional interactions occurring at the site of
seizure onset and spread may provide important insights about the pathophysiology
of seizures and aid with accurate brain mapping and resection of the epileptic focus.
In theory, removing the focus should result in a patient’s seizures being cured.
However, there is much evidence to suggest that the focus is more of a region of
seizure onset with a number of sites that can act independently to initiate seizures
(Thom et al. 2010). Seizures in animal models and in people often have a multifocal
or broadly synchronized onset. The best evidence for multifocality within the sei-
zure onset zone comes from surgical experience with intracranial monitoring. The
challenge of mapping the epileptic focus stems from the observation that the pathol-
ogy associated with focal epilepsy is often distributed across a number of brain sites
(Bertram 2009) and that current diagnostics methods frequently fall short of iden-
tifying such sites. Animal studies indicate that the neurons involved in the epileptic
circuitry have enhanced excitability throughout (Bertram et al. 1998, Mangan et
al. 2000). The implication of these observations is that each of the sites could act
independently to initiate a seizure or, potentially, to drive another site into a seizure.
Thus, in focal epilepsy, one may view a cortical region as a broad seizure onset zone,
with the potential that multiple foci can act as a seizure focus for any given seizure.
Much of our understanding about focal seizure circuitry comes from electro-
physiological recording methods. Although electrophysiology is currently the gold
standard in mapping the epileptic focus, it is often inadequate to define the boundary
of the epilepsy circuitry due to spatial sampling limitations and volume conduction.
What we truly desire is a brain-mapping modality that could give a high-resolution
real-time spatial and temporal “read out” of the dynamics of cortical processing and
seizures. It has been well established that optical contrast is highly sensitive to neu-
ronal activity (Hill and Keynes 1949, Grinvald et al. 1988). The high optical contrast
is largely due to the changes in blood volume and blood oxygenation, both of which
substantially increase during seizures, and increase in oxygenation metabolic rate
results in increased demands on autoregulatory mechanisms. Taking advantage of
the oxygenation dependence in the optical absorption spectrum of hemoglobin, opti-
cal imaging or intrinsic optical signal (IOS) methodologies have provided excellent
surface maps of epileptic focus (Haglund et al. 1992, Haglund and Hochman 2004).
Near-infrared spectroscopy (NIRS) has also shown to be able to detect seizure
(Steinhoff et al. 1996, Sokol et al. 2000, Watanabe et al. 2000). However, the major
limitation of NIRS and IOS is that each provides only surface depth information.
Greater depth information can be obtained by using tomographic reconstruction
methods such as diffuse optical tomography (Jiang 2010). Whereas diffuse optical
tomography has low spatial resolution, laser-induced PAT has both superior spatial
and temporal resolution. PAT detects absorbed photons ultrasonically by employing
the photoacoustic effect. It combines both high contrast and spectroscopic specific-
ity based on the optical absorption of both oxy- and deoxy-hemoglobin with high
ultrasonic spatial resolution. Relative to other optical imaging modalities, PAT

has the advantage of mitigating both scalp and skull light scattering by a factor of
~1,000. The end result is that PAT allows for high spatial resolution imaging of brain
at a depth considerably beyond the soft depth limit of conventional optical imaging
techniques such as confocal microscopy (Sipkins et al. 2005), two-photon micros-
copy (Denk et al. 1990), and optical coherence tomography (Huang et al. 1991). The
strong preferential optical absorption of hemoglobin makes PA imaging to have a
better imaging contrast than US; it can be difficult to visualize the microvessels with
pulse-echo US owing to the weak echogenicity (Mace et al. 2011).
In this study, PAT was employed to image seizures in an experimental acute bicu-
culline methiodide model of focal epilepsy. Bicuculline is a light-sensitive competitive
antagonist of GABAA receptors that mimics focal epilepsy when applied to brain tis-
sue. During focal application of bicuculline into the brain cortex, brains were imaged
noninvasively with an array-based PAT system. Offline, we employed measures of
brain connectivity to further identify the functional anatomy implicated in focal corti-
cal seizures. The high spatial and temporal sampling of the array-based PAT system
allowed for the first time the complete mapping of an epileptiform event in vivo. It
is also the first to map an epileptiform event at depths well below the cortex. These
findings build onto the collection of advantages of noninvasive PAT in brain dynamics
and epilepsy.
8.3.1 Methods
8.3.1.1 Animals
Male Sprague-Dawley rats (Harlan Labs, Indianapolis, IN) weighing 50g to 60g
on arrival were allowed one week to acclimate to the 12-h light/dark cycle and
given food and water ad libitum. All procedures were approved by the University
of Florida Animal Care and Use Committee and conducted in accordance with the
National Institutes of Health Guide for the Care and Use of Experimental Animals.
8.3.1.2 PAT Imaging
The 192-element array based system described in Section 4.1 was used for the in vivo
study presented in this section.
8.3.1.3 Electrode Implantation Surgery

Animals were anesthetized intraperitoneally with 1g/kg of body weight dose of
urethane. Two 300 μm-diameter stainless steel screw electrodes were implanted
within the skull for obtaining subdural multichannel cortical local field potentials
data (–0.3 mm posterior, 3 mm lateral (right) of bregma, 1 mm ventral) based on
coordinates from a rat brain atlas (Paxinos 1998). One 300 μm-diameter stainless
steel screw electrode (FHC, Bowdoin, ME) was implanted as a reference electrode
into the midline occipital bone. Cortical local field potentials were obtained using
a Tucker Davis Pentusa (Tucker Davis Technologies, Alachua, FL) neural recording
system at 12 kHz, digitized with 16 bits of resolution, and band-pass filtered from
0.5 to 6 kHz.
8.3.1.4 Induction of Seizures
Rats (n = 10) received 10 μl of 1.9 mM bicuculline methiodide and 10 μl normal
saline into the left and right parietal cortex, respectively. The infusion was performed
through the previously implanted electrode sites at a rate of 0.3 µL/min. The infu-
sion system consisted of a 100 µL gas-tight syringe (Hamilton, Reno, NV) driven by
a syringe pump (Cole-Parmer, Vernon Hills, IL) connected to polyaryletheretherk-
etone (PEEK) tubing (ID = 0.381 mm, OD = 0.794 mm, length = 0.5 m, Upchurch
Scientific, Oak Harbor, WA). The PEEK tubing was coupled to a silica cannula (ID
= 50 µm, OD = 147 µm, Polymicro Technologies, Phoenix, AZ) via a microfluidic
connector. Cortical local field potentials were recorded 5 min before each injection
and continued for up to 30 min thereafter.
8.3.1.5 Image Analysis
A custom software utility was written and incorporated into the computer software pack-
age MATLAB® (MathWorks, Massachusetts, USA) to analyze and display recorded
data. This software enabled the reconstruction of the PAT images and the determination
of the blood vessel diameter. Amira (version 5.3.3, TGS Template Graphics Software)
was used for 3D reconstruction of the seizure foci. We used the frequency-domain
Granger causality methodology to evaluate the dynamic interactions within the seizure
circuitry and the associated brain networks (Brovelli et al. 2004). The time-series data
were selected from 200 sets of PAT images sampled at 3 Hz within the seizure onset
period, and at each time point we chose nine regions of interest (ROI), R1–R9, and then
the PA signal was averaged within each ROI, as shown in Figure 8.24b. Changes in pho-
toacoustic signals were quantified as –ΔA/A in Figure 8.24b, right. The PA data shown
in Figures 8.22–8.24 were not averaged. The image color scale was determined by the
photoacoustic signal intensity in arbitrary units (between 0 and 1).

8.3.2.1 Noninvasive Epileptic Foci Localization
Noninvasive PAT image of the rat cortical vasculature (Figure 8.21a) matched well
with an anatomical atlas (Figure 8.21b) obtained after the PA imaging. The brain
structures including the middle cerebral artery, right hemispheres, left hemispheres,
left olfactory bulbs, and right olfactory bulbs are clearly shown in the PAT image.
Numbers 1–5 marked in the image indicate that the micro blood vessels with a diam-
eter of less than 100μm are also seen, which again correspond well with the rat brain
photograph. This allowed for imaging objects of 50 μm in diameter with a spatial
resolution of 150 μm (Figures 8.21a and 8.21b).
Thirty minutes following the microelectrode placement, PAT imaging was per-
formed to generate baseline tissue absorption maps, visualize micro blood vessels,
and morphological cortical landmarks (Figures 8.22b and 8.22c). Subsequently, rats
(n = 10) received 10 μl of 1.9 mM bicuculline methiodide and 10 μl normal saline
into the left [–1.0 AP, 2.5 ML, 2.4 DV] and right [–1.0 AP, 2.5 ML, 0.4 DV] pari-
etal cortex, respectively, based on coordinates from a rat brain atlas (Paxinos 1998).
Immediately following the focal infusions, PAT imaging was repeated to visual-
ize changes in the injection site tissue absorption. An increase in tissue absorption
(a) (b) MCA

MCA
RH
1 4 LH RH LH
1 4
2
5
2 5
3
3
ROB LOB
ROB LOB
2 mm
0 0.3 0.6 1
FIGURE 8.21 (a) Noninvasive PAT imaging of a rat brain in vivo with the skin and skull
intact. MCA, middle cerebral artery; RH, right hemispheres; LH, left hemispheres; LOB,
left olfactory bulbs; ROB, right olfactory bulbs. (b) Open-skull photograph of the rat brain
surface acquired after the PAT experiment. Numbers 1–5 indicate the corresponding blood
vessels in the PAT image and rat brain photograph.
was observed in the bicuculline methiodide cortical injection site and surrounding
region (Figure 8.22c, left), but not in the saline injection site or surrounding region
(Figure 8.22c, right). PAT scanning for subcortical changes was also performed.
Slices 5, 7, and 9 (Figure 8.22d) are the images obtained 3, 5, and 7 mm below the
scalp, respectively. Figure 8.22e is the 3D rendering of the epileptic foci obtained
from different tomographic layers. Each image exhibited the variable patterns of
seizure onset and propagation both at the cortical and subcortical regions. Seizures
were confirmed by concomitant time-locked PAT/video-electroencephalography
(Figure 8.22a). Experiments were repeated for each animal at 2-hour intervals.
8.3.2.2 Real-Time Monitoring of Epileptic Events

Epileptiform events were recorded with PAT from a focal region of interest of ~2 × 3
mm (Figure 8.23a). We observed a significant optical absorption change directly asso-
ciated with de-oxygenation at a wavelength of 755 nm (Figure 8.23c). Furthermore,
at the seizure onset at 1 min 6.267 s, the seizure focus measured 0.2053 mm2 and
increased to 0.5004 mm2 as the electrographic seizure time-series increased in fre-
quency and amplitude (Figure 8.23b). Corresponding rate changes (0.33 s), spike
and wave discharges, and PA images were observed in each experimental trial. The
seizure onset dynamics captured photoacoustically are best appreciated in movies
displaying how the seizure was generated and varied over time. PAT images suggest
that in addition to ictal spread from the primary focus, homotopic foci are seen in
the contralateral parietal cortex (Figure 8.24a1-4). The PA signal in the contralateral
200 μV
100 s
(a)
Max
Optical Absorption
B S BV BV
B S
2 mm
Min
(b) (c)
Max
Optical Absorption
500 μm
Min
Slice 5 Slice 7 Slice 9
(d)
x
y
z
3 mm
3 mm
(e)
FIGURE 8.22 Noninvasive epileptic foci localization. (a) EEG recordings of seizure onset
after bicuculline methiodide injection. (b) Schematic showing the location of BMI injection
and saline injection (control). B, BMI injection; S, saline injection. (c) Reconstructed PAT
image before (left) and after (right) the bicuculline methiodide injection. The BMI and saline
were injected into the left and right parietal neocortex, respectively. A significant increase
of optical absorption is seen in the region of the bicuculline methiodide injection, while no
absorption contrast is observed in the region of the saline injection relative to its surround-
ings. (d) Series of PAT images from three representative transverse planes parallel to the skin
surface with 755-nm wavelength. Slices 5, 7, and 9 are the images obtained 3, 5, and 7 mm
under the skin, respectively. The images show different spatial patterns at the seizure foci
at different tomographic layers. (e) 3D rendering of the epileptic foci from different tomo-
graphic layers.
ROI
2 mm
(a)
200 μV
20 s
(b)
65.934 Sec 1 min 6.267 Sec 1 min 6.600 Sec 1 min 6.933 Sec
0 mm2 0.2053 mm2 0.2342 mm2 0.2670 mm2
1000
500
500
1 min 7.266 Sec 1 min 7.599 Sec 1 min 7.932 Sec 1 min 8.265 Sec
0.3877 mm2 0.3986 mm2 0.4234 mm2 0.5004 mm2
500 μm
(c)
FIGURE 8.23 Real-time monitoring of ictal onset. (a) A PAT image showing the cerebral vas-
culature and the position of BMI injection. ROI, region of interest; Scale bar: 2 mm. (b) EEG
recordings showing the seizure onset about 1 min after BMI injection. (c) PAT images recording
the ictal onset in real time. At 1 min 6.267 s the area of the focus was 0.2053 mm2. The area of
the focus then increased in size over the next 2 s to 0.5004 mm2, corresponding to an increase in
the amplitude of the EEG spikes. The area of the seizure focus was derived from the PAT images
by thresholding to a pixel value one standard deviation above the pixel values from the area of
the focus during the control conditions. Scale bar: 500 μm.
1 2 3 4 1200
1000
800
600
400
200
0
200
2 mm
2 mm 400
(a)
100
50
0
50
100 R1 (focus)
R4
150
0 1 2 3
Time (s)
(b)
(c)
FIGURE 8.24 In vivo maps of contralateral homotopic seizure foci. (a) Left, the PAT image
of the rat brain; the yellow dotted rectangular shows the region of interest for analysis; Right,
noninvasive imaging of the primary and mirror foci in contralateral homotopic cortex. A1–4
shows how secondary epileptic foci were generated and diminished. The PA signal in the
secondary foci was smaller in magnitude and delayed in time compared to the signal recorded
from the primary foci. (b) Left, Granger analysis of the image around the foci and its immedi-
ate secondary foci; right, PA signal from the foci (red) and secondary area (blue) during an
ictal event. (c) Pairwise Granger analysis of the seizure foci, contralateral homotopic foci and
middle cerebral artery. C1, PAT image showing the 9 selected regions of interest; C2, arrows
indicate Granger connectivity maps between the primary and secondary foci; C3, connectiv-
ity maps between the primary and secondary foci, and the middle cerebral artery.
foci was smaller in magnitude and delayed in time compared to the signal recorded
from the primary seizure foci. We also observed a region of inverted optical signal
immediately surrounding the seizure onset zone (Figure 8.24b2-3). We performed
analysis of the optical signal at a distance of at least 2 mm away from the edge of the
focus. Our results demonstrate that the photoacoustic signal was inversely related to
the optical signal recorded from the focus (Figure 8.24b3).
To evaluate the dynamic interactions within the seizure circuitry and capture
the associated networks we used the frequency-domain Granger causality (Brovelli
et al. 2004). PAT time series data were selected from 200 sets of PAT images sam-
pled at 3 Hz within the seizure onset period. At each time point we chose 9 ROIs
(Figure 8.24b) and used the averaged optical absorption within each region for the
network analysis. Nine sets of time series analysis were classified into three groups
including the primary ictal onset area and corresponding four surrounding ROIs
including the primary or initiating seizure focus, contralateral homotopic foci, and
two ROIs surrounding the primary focus in the contralateral cortex. The middle
cerebral artery, primary seizure focus, and secondary homotopic seizure focus were
also analyzed. Results from group 1 analysis demonstrate the causal influence from
the primary focus to the surrounding regions of interest (Figure 8.24b1). The negative
zero-lag correlation (–0.68) between both the primary and secondary foci showed
that neural activities within these two regions changed in opposite direction over
time. Figure 8.24b4 shows the Granger causality analysis for group 2 data, where we
see that the primary and secondary contralateral brain foci influenced each other
mutually with a stronger influence of the primary focus relative to the contralateral
focus. Moreover, we suggest that the primary seizure focus may influence the con-
tralateral foci as previously suggested using optical intrinsic imaging in epilepsy
(Khalilov et al. 2003) and microelectrode recordings of hippocampus local field
potentials (Zhou et al. 2009). Finally, from group 3 analyses (Figure 8.24b5) we
found that the primary focus strongly influenced the hemodynamic change in the
middle cerebral artery, which in turn showed influence on the contralateral homo-
topic foci.
8.3.2.3 Dynamical Changes of Vasculature during Interictal Discharges

Figure 8.25a shows image of the rat cortex vasculature through the intact scalp
and skull. The arrow indicates the microvascularity along the direction marked
by the dashed line, representing a cross-sectional scanning using a 50 MHz ultra-
sound transducer. Positive and negative acoustic peaks induced by a 25 μm blood
vessel were sorted out as target signals to track the change of the vessel diameter.
Typical PA signals from the targeted vessel at different times show apparent ves-
sel vasomotion by ~40% during the interictal discharges (Figure 8.25b). Local
field potential recordings (Figure 8.25c) showed that interictal spikes had a strong
correlation with the discrete change in vessel size observed photoacoustically
(Figure 8.25d; error bars ± s.d. was calculated from 20 consecutive spikes).
MV
2 mm
(a)
Amplitude (A.U.)
0.2
PA Signal
0.1
0
0S
0.1 1.6 S
3.2 S
0.2 4.4 S
6.4 S
0.3
0 20 40 60 80 100 120 140 160 180
Diameter (um)
(b)
50 μY
1s
(c)
36
34
Diameter (μm)
32
30
28
26
24
22
5 10 15 20 25 30
Time (s)
(d)
FIGURE 8.25 Changes of blood vessel diameter during interictal onset. (a) PA image of
the rat cortex vasculature with the intact scalp and skull. The arrow indicates the microvas-
cularity (MV) along the dashed line direction for a cross-sectional scanning using a 50 MHz
ultrasound transducer. (b) Typical photoacoustic signals from the targeted vessel at different
times. (c) EEG recordings show the interictal spikes. (d) Change of the vessel size captured by
PAT. There was a clear correlation between the interictal spikes and the changes of the vessel
size. Error bars (±s.d.) were calculated from 20 consecutive spikes.
8.4 IMAGING OF TUMOR VASCULATURE DEVELOPMENT

Tumors growth requires sustenance by nutrients and oxygen as well as an ability to
evacuate metabolic waste and carbon dioxide (CO2). To address these needs, an
“angiogenic switch” is permanently activated, causing new vessel growth from the
adjacent host vasculature. The vasculatures within tumors are typically aberrant,
controlled by a complex biological stress that involves both the cancer cells and
stromal microenvironment. Sustained aberrant tumor angiogenesis plays a central
role in cancer carcinogenesis and metastatic potential. Several current noninvasive
imaging modalities have differing limitations for monitoring vasculature develop-
ment. For instance, X-ray computed tomography (CT) needs extrinsic contrast agent
and exposures patients to ionization radiation. Positron emission tomography (PET)
screening often involves extrinsic contrast agents. MRI is limited by its low tempo-
ral/spatial resolution. Pure high-resolution optical imaging modalities such as sin-
gle-photon, multiphoton fluorescence microscopy suffer from limited imaging depth
(<1 mm) and repeated fluorescent dye injection. Adequate noninvasive imaging can
help physicians to determine whether and when to start anti-angiogenic therapies.
In particular, such imaging is essential for monitoring the tumor response to anti-
angiogenic therapies because tumor shrinkage may not occur within a short period
of time even when anti-angiogenic treatment is effective. PAT or PAM overcomes
the limitations associated with the above-mentioned modalities and appears to be an
ideal tool for imaging/monitoring tumor vasculature development.
8.4.1 Methods
8.4.1.1 Photoacoustic Imaging System
In this study, a photoacoustic microscopy (PAM) system was used to monitor
vasculature change in breast cancer xenograft models. The system is the same as
shown in Figure 5.3. In this application, a focused ultrasound transducer (50 MHz,
3 mm aperture, and 6 mm focal length) was used to receive induced photoacous-
tic waves. PA signals amplified by two different amplifiers (one was 17 dB from
100 kHZ to 1 GHz and the other was 20 dB from 20 MHz to 3 GHz) were digi-
tized by an 8-bit data acquisition board (NI5152, National Instrument, Austin,
TX) at a sample rate of 250 MS/s. The 2D transverse scanning combined with the
depth-resolved ultrasonic detection generated 3D PA images displayed in maxi-
mum amplitude projection (MAP). Three-dimensional PA signals were processed
by Hilbert transform, normalized to the same scale (0–256) and applied with a
threshold (–6 dB level). The volumes of the blood vessels were calculated by inte-
grating the corresponding image voxels (1 for blood vessel and background being
set to 0). Entropy was calculated from normalized MAP of each PA image at the
same scale (0–256).
8.4.1.2 Mouse Breast Cancer Xenograft Models

All animals used in this study were maintained at the animal facility of the University
of Florida and handled in accordance with institutional guidelines. Athymic female
nude mice (nu/nu) at 5 to 8 weeks were purchased from Charles River Laboratories
(Charles River Laboratories, Inc., Wilmington, MA) and caged in groups of five or
fewer. Mice breast cancer xenografts were established by subcutaneous injection
of 1 × 105 NeuT (a mouse breast cancer cell line) or NeuT EMTCL2 (a malignant
variant of NeuT) cells into the mammary fat pads of the mice. Tumor volume was
calculated from caliper measurements of the large (a) and smallest (b) diameters of
each tumor using formula a × b2 × 0.4. Three days after inoculation, most tumors
had grown to ~30 mm3. All mice were euthanized when the tumor volume in the
nontreated group reached ~1000 mm3.
For in vivo siRNA (a class of antiangiogenic drugs) knockdown of the unfolded
protein response (UPR) proteins, mice with similarly sized tumors were divided into
two groups (siRNA treatment and scrambled control). Mice were anesthetized with
a mixture of ketamine (85 mg/kg)/xylazine (4 mg/kg) and were intratumoral injected
(10 μl) with scAAV2 sept-mut vectors encoding siRNAs against IRE1α or XBP-1 or
ATF6 or scrambled siRNA at a titer of 2 × 1013 genome copies per ml.

To test the feasibility of noninvasively monitoring in vivo tumor vasculature devel-
opment, we performed PA imaging for two mouse breast cancer xenograft mod-
els. In one model, mice were injected with NeuT cells, while the other model was
generated via injection of NeuT EMTCL2 cells. Serial PA imaging were carried
out on days 3, 5, 7, and 9 after tumor inoculation, respectively. The inoculation
of NeuT cells only caused a minimal tumor growth compared to a rapid tumor
growth in NeuT EMTCL2 model which tumor volume reached ~250 mm3 by day 11
(Figures 8.26a and 8.26d). Additionally, NeuT EMTCL2 cells induced a significant
vasculature development evidenced as gradual splitting large host blood vessels to
small ones. This was noticeable on day 5 and peaking on day 9. Using PA resolu-
tion and depth-section capability, we determined the kinetics of the vessel den-
sity changes surrounding tumor mass using the Entropy method (Sabuncu 2006).
Entropy is a statistical measure of randomness that can be used to characterize
the texture of the input image. In NeuT model, Entropy did not reveal a significant
increase in vessel density. However, Entropy clearly indicated that NeuT EMTCL2
inoculation resulted in increased vessel density by 20% compared with NeuT
implantation (Figure 8.26b). Another PA image extraction analysis was normalized
vessel volume changes that closely echo Entropy data. As shown in Figure 8.26c,
NeuT EMTCL2 implantation resulted in a steep increase in vessel volume, whereas
there was no significant change in tumor vessel volume detected in the NeuT model
(Figure 8.26c).
To confirm the purported anti-angiogenic activities of the siRNAs against the
UPR proteins (Figures 8.27a and 8.27b), mice of NeuT EMTCL2 xenograft mod-
els were divided into four groups including one scrambled treatment and three
siRNA treatments (IRE1α, XBP-1, and ATF6) delivered by intratumoral injection
of scAAV2 sept-mut vectors encoding the siRNAs against the UPR proteins. The
Day 3 Day 5 Day 7 Day 9
NeuT
NeuT EMTCL2
(a)
9 100
NeuT NeuT METCL2 NeuT NeuT METCL2
Normalized Volume
*p<0.05 vs NeuT *p<0.05 vs NeuT
8 80 * *
* * *
* *
Entropy
*
7 60
6 40
5
3
9
ay
ay
ay
ay
ay
ay
ay
ay
D
D
(b) (c)
300
NeuT NeuT
NeuT EMTCL2
Tumor Volume (mm3)
200
NeuT EMT CL2

100
0
0 3 5 7 9 11
Time (days)
(d)
FIGURE 8.26 In vivo serial photoacoustic imaging of developing tumor vasculature and
quantitative analysis. (a) Representative serial PA images of NeuT model (top panel) and
NeuT EMTCL2 model (bottom panel). (b) Entropy extraction for change in the vessel den-
sity over different time points as indicated. (c) Comparative analysis of normalized vessel
volumes at different time points as indicated. (d) Tumor volume was calculated from daily
caliper measurements of the large A and smallest B diameters of each tumor using formula
a × b2 × 0.4. Representative images of H&E staining for breast cancer tissue sections of NeuT
(top) and NeuT EMTCL2 (bottom) xenograft models.
Day 3 Day 5 Day 7 Day 9 Day 11
scr
IRE1
siRNA
XBP-1
ATF6
(a)
9 scr 100 scr

IRE1siRNA IRE1siRNA
XBP-1siRNA ATF6siRNA XBP-1siRNA ATF6siRNA
Normalized Volume
8 *p<0.05 vs scr, *p<0.01 vs scr 80 *p<0.05 vs scr, *p<0.01 vs scr

*
Entropy
* *
7 * * 60 *
* ** * * ** *
*
* **
*
6 ** 40 *
5 20
3
1
1
1
ay
ay
ay
ay
ay
ay
ay
ay
ay
ay
D
D
D
(b) (c)
300
scr
IRE1 siRNA
Tumor Volume (mm3)
XBP-1 siRNA
200
ATF6 siRNA
100
0
0 3 5 7 9 11
Time (days)
(c)
FIGURE 8.27 Knockdown of the siRNAs against the UPR protein resulted in decreased
tumor growth and tumor vasculature in mouse breast cancer xenografts. Mice breast can-
cer xenografts received intratumoral scAAV2 sept-mutant vector-delivered siRNAs against
IRE1α or XBP-1 or ATF6. PA imaging was performed on the same tumor inoculation site on
day 3, 5, 7, and 9 post tumor inoculations. (a) Representative PA images of different treatments:
the scrambled siRNA (top panel), IRE1α siRNA (second panel), XBP-1 siRNA (third panel)
and ATF6 siRNA (bottom panel). (b) Entropy extraction for change in the vessel density over
different time points as indicated. (c) Comparative analysis of normalized vessel volumes at
different time points as indicated. (d) Tumor volume was calculated from daily caliper mea-
surements of the large A and smallest B diameters of each tumor using formula a × b2 × 0.4.
tumor volume reached ~250 mm3 after 11 day of NeuT EMTCL2 inoculation in the
scrambled siRNAs treatment group (Figure 8.27d). Figure 8.27d also shows that
the knockdown of IRE1α and XBP-1 both exhibited significant decreased tumor
growth to a similar extent (~45% on day 11) and the ATF6 siRNA was even more
effective on inhibition of tumor growth (~54% by day 11). Serial PA imaging was
employed to monitor and evaluate the development of tumor vasculature in the
NeuT EMTCL2 xenograft models on day 3 after NeuT EMTCL2 inoculation fol-
lowed on days 5, 7, 9, and 11. Entropy method delineated a significant decrease in
vessel density (25%) by day 9 and sustained at that level thereafter in the IRE1α
siRNA treatment group compared with the scrambled group (Figure 8.27b). The
NeuT EMTCL2 xenograft models treated with the siRNAs against XBP-1 or ATF6
evidenced a steady decrease in vessel density with statistic significant in com-
parison with the scramble group (Figure 8.27b). All three siRNAs exhibited sig-
nificantly steady decrease in vessel volume compared with the scrambled siRNA
treatment (Figure 8.27c).
8.5 INTRAVASCULAR IMAGING
In the United States, the leading cause of death is acute cardiovascular events, and
the majority of them are due to the rupture of an unstable coronary atherosclerotic
plaque (Kolodgie et al. 2001, Schaar et al. 2004). The demand for imaging/diagnos-
tic technologies that can assess the vulnerability of these plaques is great. During the
past decades, there has been considerable progress in the development of new imag-
ing techniques for high-risk coronary atherosclerotic plaque detection. Compared
to the noninvasive screening modalities such as MRI, CT, and ultrasound, the cath-
eter-based intravascular imaging techniques have the advantages of providing 3D
information of structural, compositional, biochemical, and molecular features of
coronary lesions and play an increasing significant role in the detection and assess-
ment of coronary atherosclerotic plaques (Honda and Fitzgerald 2008).
Catheter-based X-ray angiography can provide real-time images of the stenosis
and is the most widely used intravascular imaging modality (Rodgers et al. 1994).
Intravascular ultrasound (IVUS) imaging offers a relatively high-resolution and
deep penetration depth cross-section image of the blood vessel (Yock et al. 1988).
Optical coherence tomography (OCT) provides very high spatial resolution images
(Toutouzas et al. 2009). Near-infrared spectroscopy (NIRS) opens up the possibil-
ity of identifying absorption properties of tissue within the artery wall (Jaffer et al.
2009). However, these techniques still suffer from various limitations when applied
to intravascular imaging. For example, X-ray angiography has the risk of unreliable
estimation of the vessel lumen narrowing when projecting the 3D arterial network to
2D silhouettes and offers no compositional information; IVUS has limited capacity
in characterizing the component of atherosclerotic plaques. OCT is limited by its 1
to 2-mm penetration depth as light loses its coherence in deeper tissue. NIRS offers
little depth/structural information of the plaque.
According to autopsy studies, researchers found that most unstable plaques are
featured a thin fibrous cap, a large lipid-rich necrotic core, and activated macrophage
near the fibrous cap (Virmani et al. 2000, Schroeder et al. 1995). In addition, there is
a strong association between atherosclerotic plaque stability and the components of
the plaques (Virmani et al. 2000, Schroeder et al. 1995). Limitations of intravascular
imaging modalities prevent them from achieving the goal of complete evaluation of
atherosclerotic plaque, which requires the 3D structural and compositional informa-
tion with deep penetration depth and high resolution. Thus, a new imaging modality/
technology that can provide deep depth, high-resolution 3D structural and composi-
tional information of the atherosclerotic plaque is highly desirable for assessment of
the vulnerability of coronary plaques.
Intravascular photoacoustic imaging (IVPA) is a catheter-based optical-induced
ultrasound imaging modality and has the potential to offer both structural and com-
positional information of the vessel wall/plaque with deep penetration depth (~1cm)
and high spatial resolution (~tens of microns). The atherosclerotic plaque is com-
posed of different chemical components that possess certain optical absorption coef-
ficients that respond to certain wavelengths of incident light. These absorbers can
be identified with the information from detected photoacoustic signals and incident
light as the ultrasound signal is proportional to the local fluence and corresponding
absorption coefficient. Specifically, by comparing the change in the photoacoustic
signals with the absorption coefficient spectra or using model-based quantitative
reconstruction algorithms (see Chapter 2), the atherosclerotic plaque components
can be characterized.
Using the PAT system described in Chapter 6 (Figure 6.12) and quantitative
PAT methods discussed in Chapter 2, experiments using ex vivo blood vessels
were conducted to demonstrate quantitative intravascular photoacoustic imaging
(IVPA). For the ex vivo experiments, IVPA measurements at multiple wavelengths
(710, 800, 900, 1100, and 1150 nm) were performed to extract the composition
of plaques. Figure 8.28 shows the reconstructed images by the delay-and-sum
(Figure 8.28c) and by the FE-based method (Figure 8.28d) for one set of repre-
sentative ex vivo atherosclerotic blood vessel (Figure 8.28a gives the photograph
of the sample). Comparing these results with the artery sample, we can see that
the plaque shows up in each image at all wavelengths and in the right position
while the shape and PA signal or absorption coefficient change in response to the
wavelength.
Using the absorption coefficient images at the five wavelengths coupled with the
Beer-Lambert law (Figure 8.28b), we can obtain the concentrations of the four main
components of blood vessels, which are lipid, elastin, collagen, and water. Based
on these compositional images shown in Figure 8.29, the wet weight percentages of
the components in this atherosclerotic artery were estimated as 15, 28, 32, and 25%
for lipid, elastin, collagen, and water, respectively. Comparing with normal tissue
which has 24% elastin, 37% collagen, and 1.6% lipid (Black and Hastings 1998), our
reconstructed results show a higher lipid percentage, which means that a plaque with
a lipid core existed in the imaging section.
Absorption Coefficient (cm1) Hb Water 15

HbO2 Intima 10 0.08
102 Fat Media
Fatty acid Adventitia 5 0.06
10 0 0
0.04
5
102 10 0.02
15
800 1000 1200 10 0 10
Wavelength
(b)
15 0.05 15
10 0.04
0.04 10
5 5 0.03
0.03
0 0
0.02 0.02
5 5
10 0.01 10 0.01
15 15
10 0 10 10 0 10
(d) 15 15
10 0.025 10 0.02
5 0.02 5
0.015
0 0.015 0
5 5 0.01
0.01
10 10 0.005
0.005
15 15
10 0 10 10 0 10
FIGURE 8.28 (a) Photograph of a human atherosclerotic artery sample; (b) optical absorp-
tion coefficient spectra of various components within the artery; (c) PA images recovered by
the delay-and-sum method at multiple wavelengths; (d) Absorption coefficient images recov-
ered by the FE-based method at multiple wavelengths.
8.6 BREAST IMAGING
Breast cancer is a leading cause of death for women in the United States. The best
way of combating the increased incidence of breast cancer is early detection. While
X-ray mammography has long been accepted as the most effective method to detect
breast cancer prior to outward signs and symptoms, it has unacceptable false nega-
tive rates for patients with radiodense breast tissue. For example, the sensitivity of
screening mammography decreases from a high of 88% in fatty breast tissue to 62%
in dense breast tissue (Carney et al. 2003). Patients with dense breasts represent the
general population of premenopausal women as well as those with fibrocystic tissue
15 15
1
10 0.3 10
0.8
5 5
0.2 0.6
0 0
5 5 0.4
0.1
10 10 0.2
15 15
10 0 10 10 0 10
Lipid 15% Elastin 28%
15 15
1.2
10 10 0.8
1
5 5
0.8 0.6
0 0
0.6 0.4
5 0.4 5
10 10 0.2
0.2
15 15
10 0 10 10 0 10
Collagen 32% Water 25%
FIGURE 8.29 Reconstructed ex vivo lipid, elastin, and collagen images.
disease. Yet, cancer in younger women tends to be more virulent and grow faster.
Another concern is that the positive predictive value of X-ray mammography is quite
low. This has numerous ramifications both medical and economic, in terms of the
number of necessary biopsies. Consequently, there is a critical need to develop breast
cancer detection methods that could serve either a complementary or competitive
role with respect to conventional X-ray mammography.
Several other conventional techniques are currently under investigation for breast
cancer detection, including ultrasound (US), MRI, CT, positron emission mammog-
raphy (PEM), and breast-specific gamma imaging (BSGI). Among these conventional
alternatives or supplements, MRI, PEM, and BSGI are the most promising solutions to
dense breast imaging. Bluemke and colleagues showed a sensitivity of 88% and cor-
responding specificity of 68% regardless of breast density from an MR study of 821
women (Bluemke et al. 2004). Similar high sensitivities of MR for breast cancer have
also been found in other studies involving targeted imaging of high-risk populations
(Warner et al. 2004). While MR appears to offer high sensitivity in the dense breast,
a major disadvantage of MRI is the limited specificity resulting from its enhancement
of benign breast lesions (Kuhl 2007). Recent emerging nuclear imaging techniques
like PEM and BSGI provide molecular information of tumor tissue. Carmen and col-
leagues showed 88% sensitivity and speciﬁcity from PEM of 808 subjects (Carmen et
al. 2005). However, the high cost and the use of exogenous contrast agents may render
these techniques impractical for frequent monitoring and screening.
The need for a nonionizing, noninvasive imaging technique that can provide
high-contrast images at microscale resolution with functional information for breast
cancer can be fulfilled with PAT. As described in Chapter 2, PAT can obtain infor-
mation on hemoglobin and oxygen saturation at high resolution and contrast, without
the use of exogenous contrast agents, which is a significant advantage when com-
pared with other tumor hypoxia imaging techniques (e.g., blood oxygen level depen-
dent-MRI and PET).
Another unique advantage of PAT for breast imaging is to facilitate monitoring
treatment for breast cancer. At the time of diagnosis of breast cancer, more than 30%
of cases are locally advanced breast cancer (LABC). For these women, lumpectomy,
or curative surgery, is often an option if preoperative (neoadjuvant) chemotherapy
(NAC) is given to shrink the tumor to allow for optimal surgical therapy. While NAC
is emerging as a promising treatment strategy for locally advanced breast cancer,
deploying monitoring methods during the course of therapy is important for mak-
ing further advances. MR and fluorodeoxyglucose (FDG) PET have been evaluated
in several clinical trials and have proven useful in this setting, but both come at
considerable cost and inconvenience (Partridge et al. 2005, Rousseau et al. 2006).
Assessing tumor shrinkage by clinical breast examination is possible as well but
is poorly correlated with treatment responses. Nonionizing, noninvasive PAT with
high resolution has shown its promise in NAC monitoring (see Figures 8.33 and
8.34). Significant changes in absolute total hemoglobin (HbT) and oxygen saturation
(O2%) of tumor regions have been observed during NAC from patient data. Our ini-
tial results have shown that PAT is very promising and invaluable in assessing and
predicting tumor response to NAC of breast cancer.
8.6.1 Photoacoustic Tomography System

A prototype PAT system for breast cancer detection in humans is shown in
Figure 8.30. It can produce cross-section images of breast tissue with a ring-shaped
homemade array of 64 single-layer acoustic transducers (Xi et al. 2012). A scanning
(a)
Data
Breast Amplifier Acquisition
Boadrs
(b) Computer
Light Delivery
System
Pulsed
Mirror Laser
(c)
FIGURE 8.30 (a) Photograph of breast/transducer interface. (b) Photograph of PAT clinical
exam table. (c) Schematic of PAT system.
light delivery system was built to illuminate a large area of breast tissue for opti-
mized tissue penetration. Frequency response of the transducers is from 380 kHz
to 1.48 MHz and maximum frequency response is up to 2 MHz. The PAT system
offers a spatial resolution of 0.5 mm. A depiction of the breast/transducer interface
and the system in a clinical setting is shown in Figures 8.30a and 8.30b, respectively.
During the exam, the patient was in a prone position with her breast pendant through
an aperture. The breast was slightly compressed between a PETG plate at the cranial
plane and the ultrasound detector array at the caudal plane surrounding the breast.
Cross-section images of breast tissue can be captured by a ring-shaped interface. The
breast was illuminated through the PETG plate with light from a pulped Ti:Sapphire
laser optically pumped with a Q-switched Nd:YAG laser.
8.6.2 Clinical Experiments
Here we demonstrate two pilot clinical studies with our prototype system shown
in Figure 8.30. Two studies including breast cancer detection and monitoring NAC
treatment for breast cancer are demonstrated below. The first study involved breasts
with different types of lesions, diagnostic imaging (mammogram, MRI, and/or
ultrasonography), and healthy controls. In the second part, a case study is shown to
monitor NAC response using PAT technique on a patient (47-year-old female) who
underwent NAC prior to surgery. Maps of absolute total hemoglobin (HbT = HbR
+ HbO2) and oxygen saturation (O2%) of coronal plane were calculated from two
absorption coefficient images reconstructed at 775 and 808nm, respectively, for both
studies. This is the first attempt to use functional PAT methods in clinical breast can-
cer diagnosis and monitoring NAC. Protocols for both studies were approved by the
institutional review board and conducted in full compliance with the accepted stan-
dards for research involving humans. Written informed consent was obtained from
all the participants. Both African American and Caucasian patients were included
in the study.
8.6.2.1 PAT for Breast Cancer Detection

From all six cases including three healthy and three malignant breasts, the average
HbT is 21.6 ± 4.5 µM and O2% is 67.5 ± 2.2% which is compatible with findings of
studies using other optical methods (Srinivasan et al. 2003). One case from ductal
carcinoma in situ (DCIS) and one case from infiltrating ductal carcinoma (IDCA) is
shown below.
Case 1 is a 48-year-old woman with high-grade IDCA (Nottingham Grade 8) and
maximum linear dimension of IDCA in a single core of 1.2 cm was reported from
pathology. Ultrasound images acquired two weeks before PAT examination of lesion
(Figure 8.31a) indicated a 1.9 × 1.1 × 1.0 cm mass. HbT map of coronal plane from
PAT examination is shown in Figure 8.31b with increased HbT concentration in the
lesion area of size 1.1 × 0.8 cm, which is compatible with imaging and pathologic
findings. Inhomogeneous circular pattern in the tumor region is also easily recog-
nized in O2% map of coronal plane shown in Figure 8.31b. Low oxygen saturation is
an indication of tumor hypoxia or necrosis, and several investigators have reported
a relationship between hypoxic tumors and lower partial pressure of oxygen (Denko
(a) US (b) HbT O2%

(μM) (%)
1
40
0.8
30 0.6
20 0.4
0.2
10
0
10 mm 50 mm
FIGURE 8.31 (a) Ultrasound image of lesion area. (b) HbT and O2% maps of coronal plane
from PAT.
2008). The finding of increased oxygen saturation in part of the tumor region is con-
firmed with other optical methods (Srinivasan et al. 2003).
Case 2 is a PAT study with a malignant lesion in the left breast while no abnormal-
ity in the right breast from pathology report and diagnostic imaging (mammogram,
ultrasound, and/or MRI). The left breast lesion is high-grade DCIS and invasive
mammary carcinoma with apocrine features and squamous metaplasia (Nottingham
Grade 7). Contrast MRI examination with gadolinium was taken one week before
PAT examination. Three-dimensional MRI images of both breasts shown in
Figures 8.32c and 8.32d were constructed with 2D slices obtained from the MRI
(a) (b)
Right Breast (Healthy) Left Breast (Abnormal)
(µM) (%)
1
40
0.8
PAT 30 0.6
20 0.4
10 0.2
0
50 mm HbT O2% HbT O2 % HbT O2
(c) R L (d)
MRI 50
mm
FIGURE 8.32 (a) MRI image of healthy breast. (b) MRI image of abnormal breast.(c) HbT
and O2% maps of coronal plane from PAT of healthy breast. (d) HbT and O2% maps of coronal
plane from PAT of abnormal breast.
scans. It showed benign MRI in right breast and abnormal left breast—an irregular
area of suspicious non-masslike enhancement with plateau kinetics that extends 5 cm
in AP dimension and 2.6 cm in craniocaudal dimension, and the anterior margin of
the suspicious area of enhancement begins 5.6 cm posterior to the nipple. HbT and
O2% maps of coronal plane taken from PAT are shown in Figures 8.32a and 8.32b
for the right and left breast, respectively. We can see that our multiwavelength PAT
system successfully detected a lesion area at 5.6 cm depth from nipple. Increased
HbT concentration is detected with diameter 1.9 × 2.7 cm (maximum diameter in
each axis), which is compatible with the MRI findings. Decreased O2% concentra-
tion circling an increased concentration is shown in O2% map (Figure 8.32b), which
may relate with blood flow and necrosis of tumor region. From the right breast PAT
images shown in Figure 8.32a, we can see that while some nonhomogeneous pattern
exists in HbT and O2% maps, the contrast of these benign patterns is less than 1.2
for both HbT and O2%, which is much smaller than the contrast of the lesion region
compared to normal tissue (which is higher than 1.6). Similar findings are confirmed
with other in vivo cases examined by this PAT system. These findings indicate that
PAT has a great potential for breast cancer diagnosis, and its further clinical explora-
tion is very promising.
8.6.2.2 PAT for Neoadjuvant Monitoring

The study subject (47-year-old female) was referred for NAC and has high grade
DCIS and invasive mammary carcinoma with apocrine features and squamous
metaplasia in left breast as indicated in the initial pathology report before NAC.
Nottingham histologic score is 7 with ER–, PR+, and Her2–. After initial evaluation,
the patient received NAC which included four cycles of dose-dense Adriamycin plus
cyclophosphamide and then four cycles of paclitaxel given every 2 weeks. In the
PAT study of NAC treatment monitoring, the first scan took place before the start
of treatment to obtain information as baseline. Subsequent scans took place every
month after the start of NAC to monitor tumor response to the treatment. The final
scan was scheduled after completion of chemotherapy before surgery. Both HbT and
O2% maps were obtained from multiwavelength PAT absorption images.
Figure 8.33a shows HbT and O2% maps obtained at baseline, each month during
NAC treatment and after treatment before surgery. The images clearly indicate the
dynamic change of HbT and O2% concentration in the tumor region from baseline
till the end of NAC treatment. Quantitative analysis is shown in Figure 8.34. The
HbT and O2% of tumor region responded dramatically after 1 month of NAC treat-
ment from 217.8 to 90.8 mm2 measured by FWHM from the HbT maps, then shrank
consistently from 90.8 to 39.9 mm2. Contrast of lesion area of HbT and O2% is con-
tinuously decreasing from 2.6 to 1.7 and 1.3 to 1.1, respectively, which related well
with the dynamic changes of the tumor. Detailed comparison with contrast MRI and
pathology of prior NAC and after NAC before surgery are discussed below.
The initial pretreatment contrast MRI was taken 1 week before PAT exam and
the initial US was taken 2 weeks before PAT exam. The US image of the lesion
area and an MRI image of the coronal plane (5.6 cm from nipple) reconstructed
from the original 2D horizontal slices are shown in Figure 8.33b. Targeted ultra-
sound demonstrated a multicystic lesion located at 3 o’clock. MRI image shows
Before NAC 1 month 2 months 3 months After NAC

(a)
40
30
HbT
20
10
20 mm
1
0.8
O2% 0.6
0.4
0.2
0
20 mm 20 mm
MRI US US MRI
(b) (c)
FIGURE 8.33 (a) HbT and O2% maps of an invasive mammary carcinoma with DCIS patient
obtained at baseline, each month during treatment and after treatment. (b) Reconstructed cor-
onal plane (5.6cm from nipple) from MRI and US image of lesion area before NAC.
(c) Reconstructed coronal plane (5.6cm from nipple) from MRI and US image of lesion area
after NAC. Note that all the PAT, MRI, and US images displayed in the same scale.
Contrast 3 250 Area
2.5
200
2
150
1.5 HbT contrast
100 O2% contrast
1
Area(mm2)
50
0.5
0 0
B NAC 1M 2M 3M A NAC
FIGURE 8.34 Quantitative analysis of size of lesion area, contrast of HbT and O2% maps of
tumor and surrounding normal tissue from PAT exams before NAC, 1 month, 2 months, and
3 months after first NAC treatment till NAC treatment was completed before surgery.
an irregular area of suspicious non-masslike enhancement with plateau kinetics of

1.8 × 2.3cm (Figure 8.33b). HbT and O2% maps of coronal plane from PAT are
shown in Figure 8.33a. Increased HbT concentration and oxygen saturation in the
tumor region are detected with a size of 1.9 × 2.7 cm, which is compatible with
MRI findings. The shape of the increased HbT and O2% region follows closely to
the tumor region depicted by MRI. This is not surprising as the contrast MRI pri-
marily images the tumor vasculature. While the MRI image shows two separate
tumor regions (Figure 8.33b), the HbT image provides more fine detail of the tumor
vasculature (a diameter of 2.1 mm of increased HbT is measurable from the bottom
part of the tumor, Figure 8.33a), and the edge of the tumor in the O2% image shows
decreased O2% concentration indicating tumor hypoxia or necrosis at the boundary.
Both MRI and US images were taken after NAC and before surgery, as shown
in Figure 8.33c. A 1.3 × 0.4 cm heterogeneous, hypoechoic mass that likely con-
tains microcalcifications was reported from US. A 0.6 × 0.8 cm region of non-
masslike enhancement is measured from MRI (Figure 8.33c). Pathology reported
an invasive carcinoma with the largest focus of 0.4 cm, and surrounding DCIS
was present at surgery. From O2% maps reconstructed from the last exam with
PAT, the lesion area is hardly detectable with contrast less than 1.2. Increased HbT
concentration from coronal plane is detected with a size of 0.7 × 0.6 cm, which
should be contributed from IDCA. This finding is compatible with the MRI and
US diagnosis report.
Bibliography
R. Acar, and C. R. Vogel, “Analysis of bounded variation penalty methods for ill-posed prob-
lems,” Inverse Probl. 10, 1217–1229 (1994).
F. Akasheh, T. Myers, J.D. Fraser, S. Bose, and A. Bandyopadhyay, Sens. Actuators, A 111,
275–287 (2004).
M. Akiyama, T. Kamohara, K. Kano, A. Teshigahara, Y. Takeuchi, and N. Kawahara, Adv.
Mater. 21, 593–596 (2009).
H. An, W. Lin, “Cerebral venous and arterial blood volumes can be estimated separately in
humans using magnetic resonance imaging,” Magn. Reson. Med, 48, 583–588 (2002).
V.G. Andreev, A.A. Karabutov and A.A. Oraevsky, “Detection of ultrawide-band ultrasound
pulses in optoacoustic tomography, ” IEEE Trans. Ultrason. Ferroelectr. Freq. Control
50, 1383–1390 (2003).
ANSI/IEEE Std 176-1987, IEEE Standard on Piezoelectricity, The Institute of Electrical and
Electronics Engineers, Inc., New York.
L. Assersohn, T. J. Powles, S. Ashley, A. G. Nash, A. J. Neal, N. Sacks, J. Chang, U. Querci
della Rovere, and N. Naziri, “Local relapse in primary breast cancer patients with unex-
cised positive surgical margins after lumpectomy, radiotherapy and chemoendocrine
therapy,” Ann Oncol. 10(12), 1451–1455 (1999).
E.D. Aydin, C.R.E. de Oliveira, and A.J.H. Goddard, “A finite element-spherical harmonics
radiation transport model for photon migration in turbid media,” J. Quant. Spectrosc.
Radiat. Transfer 84, 247–60 (2004).
G. C. Balch, S. K. Mithani, J. F. Simpson, and M. C. Kelley, “Accuracy of intraoperative
gross examination of surgical margin status in women undergoing partial mastectomy
for breast malignancy” Am Surg. 71(1), 22–27 (2005).
J. Barton, T. Pfefer, A. Welch, “Optical Monte Carlo modeling of a true port wine stain anat-
omy,” Opt. Expr. 2, 391–396 (1998).
K.J. Bath, Numerical methods in finite element analysis, Prentice Hall, Englewood Cliffs, New
Jersey (1976).
A.G. Bell, Am. J. Sci. 20, 305 (1880).
A.T. Berg, J. Langfitt, S. Spencer, B. Vickrey, “Stopping antiepileptic drugs after epilepsy surgery
A survey of US epilepsy center neurologists,” Epilepsy Behav 10 (2), 219–222 (2007).
J. Bergen, H.A. von Recum, H.F. von Recum, T.T. Goodman, A.P. Massey, S.H. Pun, “Gold
nanoparticles as a versatile platform for optimizing physicochemical parameters for tar-
geted drug delivery,” Macromolecular Bioscience 6, 506–516 (2006).
R. Berkow, M. Beers, R. Bogin, A. Fletcher, The Merck Manual of Medial Information (Home
Edition). Whitehouse Station, New Jersey: Merck Research Laboratories, 1997. Chap.
A. III, “Common Medical Tests,” 1375–1376.
E.H. Bertram, D. Zhang, P. Mangan, N. Fountain, D. Rempe, “Functional anatomy of limbic
epilepsy: a proposal for central synchronization of a diffusely hyperexcitable network,”
Epilepsy Res. 32 (1–2), 194–205 (1998).
E.H. Bertram, “Temporal lobe epilepsy: where do the seizures really begin?” Epilepsy Behav.
14 Suppl 1, 32–37 (2009).
J. Beuthan, V. Prapavat, R. Naber, O. Minet, G. Muller, “Diagnostic of inflammatory rheu-
matic diseases with photon density waves”, Proc. SPIE 2676, 43–53 (1996).
J. Black, G. Hastings, Handbook of biomaterial properties, Chapman & Hall, 1998.
F.J. Blanco, M.J. Lopz-Armada, E. Maneriro, “Mitochondrial dyfunction in osteoarthritis,”
Mitochondrion 4: 715–728 (2004).
273
274 Bibliography
D.A. Bluemke, C.A. Gatsonis, M.H. Chen, et al., “Magnetic resonance imaging of the breast
prior to biopsy,” JAMA 292 (22): 2735–2742 (2004).
G. Bonadonna, P. Valagussa, C. Brambilla, L. Ferrari, A. Molinterni, M. Terenziani, M.
Zambetti, “Primary chemotherapy in operable breast cancer: eight years experience at
the Milan Cancer Institute,” J Clin Oncol 1998.
K.D. Brandt, “Pathogenesis of osteoarthritis,” in Textbook of Rheumatology (W.N. Kelly,
E.D. Harris, S. Ruddy, C. Sledge, editors), fifth edition, WB Saunders, Philadelphia,
1997.
M. Brittberg, A. Lindahl, A. Nilsson, C. Ohlsson, O. Isaksson, L. Peterson, “Treatment of deep
cartilage defects in the knee with autologous chondrocyte transplantation,” New Engl. J.
Med. 331, 879–895 (1994).
A. Brovelli, M. Ding, A. Ledberg, Y. Chen, R. Nakamura, S. L. Bressler, “Beta oscillations in
a large-scale sensorimotor cortical network: directional influences revealed by Granger
causality,” Proc Natl Acad Sci U S A 101, 9849–9854 (2004).
R. Carmen, M. Renee, M. D. Blaufox, “A meta-analysis of FDG-PET for the evaluation of
breast cancer recurrence and metastases,” Breast Cancer Research and Treatment 90,
105–112 (2005).
P.A. Carney, D.L. Miglioretti, B.C. Yankaskas, et al. “Individual and combined effects of age,
breast density, and hormone replacement therapy use on the accuracy of screening mam-
mography,” Annals of Internal Medicine 138 (3), 168–175 (2003).
S. Carp, J. Selb, “Experimental and theoretical studies of oxygen gradients in rat pial micro
vessels,” Opt Express, 16, 16064–16078 (2008).
K.M. Case, and P.F. Zweifel, Linear Transport Theory, Addison-Wesley, Reading (1967).
W.P. Chan, P. Lang, M.P. Stevens, “Osteoarthritis of the knee: comparison of radiology, CT,
and MR imaging to asses extent and severity,” Am. J. Res. 157, 799–806 (1991).
D.E. Dausch, J.B. Castellucci, D.R. Chou, and O.T. von Ramm, IEEE Trans. Ultrason.
Ferroelectr. Freq. Control 55, 2484–2492 (2008).
A. De la Zerda, J. Kim, E. Galanzha, S. Cambhir, V. Zharov, “Advanced contrast nanoagents
for photoacoustic molecular imaging, cytometry, blood test and photothermal theranos-
tics,” Contrast Media Mol. Imaging 6, 346–369 (2011).
H. Dehghani, D.T. Delpy, and S.R. Arridge, “Photon migration in non-scattering tissue and the
effects on image reconstruction,” Phys. Med. Biol. 44, 2897–2906 (1999).
S. DelVecchio, M.P. Stoppelli, M.V. Carriero, R. Fonti, O. Massa, P.Y. Li, G. Botti, M. Cerra,
G. D’Aiuto, G. Esposito, M. Salvatore. Cancer Res. 53, 3198–3206 (1993).
W. Denk, J.H. Strickler, and W.W. Webb, “Two-photon laser scanning fluorescence micros-
copy,” Science 248: 73–76 (1990).
N.C. Denko, “Hypoxia, HIF1 and glucose metabolism in the solid tumour,” Nature Rev.
Cancer 8, 705–713 (2008).
C. Desjardins, B. Duling, “Microvessel hematocrit: measurment and implications for capillary oxy-
gen transport,” Am. J. Physiol. (Heart and Circulatory Physiology) 252, H494–H503 (1987).
D.C. Dobson, and F. Santosa, “An image-enhancement technique for electrical impedance
tomography,” Inverse Probl. 10, 317–334 (1994).
D.C. Dobson, and F. Santosa, “Recovery of blocky images from noisy and blurred data,” SIAM
J. Appl. Math. 56, 1181–1198 (1996).
E. Dublin, A. Hanby, N.K. Patel, R. Liebman, D. Barnes, Am. J. Pathol. 157, 1219–1227 (2000).
J.J. Duderstadt, W.R. Martin, Transport theory, John Wiley & Sons, New York (1979).
T. Duong, S. Kim, “in vivo MR measurements of regional arterial and venous blood volume
fractions in intact rat brain,” Magn. Reson. Med., 43, 393–402 (2000).
J. Engel, “Surgical treatment for epilepsy: Too little, too late?” JAMA 300 (21), 2548–2550
(2008).
S.A. Ermilov, T. Khamapirad, A. Conjusteau, M.H. Leonard, R. Lacewell, K. Mehta, T. Miller,
A.A. Oraevsky, J. Biomed. Opt. 14 (2), 024007 (2009).
Bibliography 275
R. Fisher, W. van Emde Boas, W. Blume, C. Elger, P. Genton, P. Lee, J. Engel, “Epileptic
seizures and epilepsy: definitions proposed by the International League Against
Epilepsy (ILAE) and the International Bureau for Epilepsy (IBE),” Epilepsia 46 (4),
470–472 (2005).
M. Firbank, S.R. Arridge, M. Schweiger, D.T. Delpy, “An investigation of light transport through
scattering bodies with non-scattering region,” Phys. Med. Biol. 41, 767–783 (1996).
M. Fukuda, M. Nishihara, K. Imano, Jpn. J. Appl. Phys. 45, 4556 (2006).
E.I. Galanzha, E.V. Shashkov, T. Kelly, J.W. Kim, L. Yang, V.P. Zharov, Nat. Nanotech. 4,
8558–60 (2009).
N. Gerwin, C. Hops, A. Lucke, “Intraarticular drug delivery in osteoarthritis,” Advanced Drug
Delivery Reviews 58, 226–242 (2006).
J. Greenfield, J. Geyer, P.R. Carney, Reading EEGs: A Practical Approach. Lippincott Wilkins
Raven, Philadelphia (2011).
J.F. Greenleaf, R.C. Bahn, “Clinical imaging with transmissive ultrasound computerized
tomography,” IEEE Trans. Biomed. Eng. 28, 177–185 (1981).
A. Grinvald, R.D. Frostig, E. Lieke E, et al. “Optical imaging of neuronal activity,” Physiol
Rev 68,1285–366 (1988).
S.R. Grobmyer, B. Moudgil, Cancer Nanotechnology, Humana Press, 2010.
X. Gu, Y. Xu, H. Jiang, “Mesh-based enhancement schemes in diffuse optical tomography”,
Med. Phys. 30, 861–869 (2003).
R.O. Guldiken, J. Zahorian, F.Y. Yamaner, F.L. Degerteken, IEEE Trans. Ultrason. Ferroelectr.
Freq. Control 56 (6), 1270–1276 (2009).
V.E. Gusev, A.A. Karabutov, Laser optoacoustics, AIP, New York (1993).
M.M. Haglund, G.A. Ojemann, D. Hochman, “Optical imaging of epileptiform and functional
activity from human cortex,” Nature 358, 668–671 (1992).
M.M. Haglund, D. Hochman, “Optical imaging of epileptiform activity in human neocortex,”
Epilepsia 45, 43–47 (2004).
G. Hale, M. Querry, “Optical constants of water in the 200 nm to 200 µm wavelength region,”
Appl. Opt. 12, 555–563 (1973).
M.G. Harisinghani, J. Barentsz, P.F. Hahn, W. Deserno, S. Tabatabaei, C. van de Kaa, J. de la
Rosette, R. Weissleder, “Noninvasive detection of clinically occult lymph–node metas-
tases in prostate cancer,” N. Engl. J. Med. 348 (25), 2491–2499 (2003).
W.A. Hauser, L.T. Kurland, “The epidemiology of epilepsy in Rochester, Minnesota, 1935
through 1967,” Epilepsia 16 (1), 1–66 (1975).
W.A. Hauser, “Seizure disorders: the changes with age,” Epilepsia 33 Suppl 4, S6–14 (1992).
A. Hemsen, L. Riethdorf, N. Brunner, J. Berger, S. Ebel, C. Thomassen, F. Jänickeand K.
Pantel. Int. J. Cancer 107, 903–909 (2003).
Y. Henrotin, B. Kurz, T. Aigner, “Oxygen and reactive oxygen species in cartilage degradation:
friends or foes,” Osteoarthritis and Cartilage 13, 643–654 (2005).
L.G. Henyey, and J.L. Greenstein, “Diffuse radiation in the galaxy,” Astro. J. 93, 70–83 (1941).
M. Hernandez, R. Brennan, G. Nowman, “Cerebral blood flow autoregulation in the rats,”
Stoke 9, 150–154 (1978).
A.H. Hielscher, R.E. Alcouffe, R.L. Barbour, “Comparison of finite-difference transport and
diffusion calculations for photon migration in homogeneous and heterogeneous tissues,”
Phys. Med. Biol. 43, 1285–1302 (1998).
D.K. Hill, R.D. Keynes, “Opacity changes in stimulated nerve,” J Physiol 108, 278–281 (1949).
C.G.A. Hoelen, F.F.M. de Mul, “Image reconstruction for photo-acoustic scanning of tissue
structures,” Appl. Opt. 39, 5872–5883 (2000).
W.J. Hoskins, “Prospective on ovarian cancer: why prevent?” J Cell Biochem Suppl. 23,189–
199 (1995).
D.S. Howell, R.D. Altman, “Cartilage repair and conservation in osteoarthritis”, Rheum. Dis.
Clin. North Am.19, 713–724 (1993).
276 Bibliography
D. Huang, E.A. Swanson, C.P. Lin, J.S. Schuman, W.G. Stinson, W. Chang, M.R. Hee, T.
Flotte, K. Gregory, C.A. Puliafito, J.G. Fujimoto, “Optical coherence tomography,”
Science 254, 1178–1181 (1991).
Y.L. Huang, X.F. Zhuang, E.O. Hæggstrom, A.S. Ergun, C.H. Cheng, B.T. Khuri-Yakub, IEEE
Trans. Ultrason. Ferroelectr. Freq. Control 56 (1), 136–145 (2009).
N. Iftimia, H. Jiang, “Quantitative optical image reconstruction of turbid media by using
direct-current measurements,” Appl. Opt. 39, 5256–5261 (2000).
S.A. Itay, A. Abramovici, Z. Zero, “Use of cultured embryonal chick epiphyseal chrondrocytes
as grafts for detects in chick articular cartilage”, Clin. Orthop. 220, 284–303 (1987).
J. A. Jacobson, G. Girish, Y. Jiang, B.J. Sabb, “Radiographic evaluation of arthritis: degenera-
tive joint disease and variations,” Radiology 248, 737–747 (2008).
F.A. Jaffer, P. Libby, R. Weissleder, Optical and multimodality molecular imaging: insights
into atherosclerosis,” Arterioscler Thromb Vasc Biol 29, 1017–1024 (2009).
A. Jemal, R. Siegel, E. Ward, Y. Hao, J. Xu, T. Murray, M.J. Thun, “Cancer statistics 2008. CA
Cancer,” J Clin. 58 (2), 71–96 (2008).
C.H. Jeon, J.K. Ahn, J.Y. Chai, H.J. Kim, E.K. Bae, S.H. Park, E.Y. Cho, H.S. Cha, K.S. Ahn, E.M.
Koh, “Hypoxia appears at pre-arthritic stage and shows co-localization with early synovial
inflammation in collagen induced arthritis,” Clin. Exp. Rheumatol. 26, 646–648 (2008).
H. Jiang, K.D. Paulsen, U.L. Osterberg, M.S. Patterson, “Frequency-domain optical image
reconstruction in heterogeneous media: an experimental study of single-target detect-
ability,” Applied Optics 36, 52–63 (1997).
H. Jiang, K.D. Paulsen, U.L. Osterberg, M.S. Patterson, “Frequency-domain near-infra-
red photo diffusion imaging: initial evaluation in multi-target tissue-like phantoms”,
Medical Physics 25, 183–193 (1998).
H. Jiang, N.V. Iftimia, Y. Xu, J.A. Eggert, L.L. Fajardo, K.L. Klove, “Near-infrared optical imag-
ing of the breast with model-based reconstruction, ” Acad. Radiol. 9 (2), 186–194 (2002).
H. Jiang, Diffuse optical tomography: principles and applications, CRC Press (November 2010),
H. Jiang, Z. Yuan, X. Gu, “Spatially varying optical and acoustic property reconstruction using
finite element-based photoacoustic tomography,” J. Opt. Soc. Am. A, 23, 878–888 (2006).
J.M. Jin, The finite element method in electromagnetics, John Wiley & Sons, New York, 2002.
G. Kanschat, “A robust finite element discretization for radiative transfer problems with scat-
tering,” East-West J Num. Math, 6, 265–272 (1998).
I. Khalilov, G.L. Holmes, Y. Ben-Ari, “in vitro formation of a secondary epileptogenic mirror
focus by interhippocampal propagation of seizures,” Nat Neurosci 6, 1079–1085 (2003).
A. Kienle, L. Lilge, A. Vitkin, M. Patterson, B. Wilson, R. Hibst, R. Steiner, “Why do veins
appear blue? A new look at an old question,” Appl. Opt. 35, 1151–1160 (1996).
A.D. Klose, U. Netz, J. Beuthan, A.H. Hielscher, “Optical tomography using the time-inde-
pendent equation of radiative transfer: Part 1. Forward model,” J. Quant. Spectrosc.
Radiat. Tranfer 72, 691–713 (2002).
G.C. Knollman, J L.S. Bellin, J.L. Weaver, J. Acoust. Soc. Am. 49, 253–261 (1970).
L. Knott, A Bailey, “Collagen cross-links in mineralizing tissues: a review of their chemistry,
function and clinical relevance,” Bone 22, 181–187 (1998).
R. Kollkman, W. Steenbergen, T.G. van Leeuwen, “Reflection mode photoacoustic measure-
ment of speed of sound,” Opt. Express, 15, 3291–3300 (2007).
F.D. Kolodgie, A.P. Burke, A. Farb, et al. “The thin-cap fibroatheroma: a type of vulnerable
plaque: the major precursor lesion to acute coronary syndromes.” Curr Opin Cardiol 16,
285–292 (2001).
S. Kommareddy, M. Amiji, “Biodistribution and pharmacokinetic analysis of long-circulating
thiolated gelatin nanoparticles following systemic administration in breast cancer-bear-
ing mice,” J Pharm Sci 96 (2), 397–407 (2007).
K.P. Kostli, M. Frenz, H. Bebie, H.P. Weber, “Temporal backward projection of optoacoustic pres-
sure transient using Fourier transform method,” Phys. Med. Biol., 46, 1863–1872 (2000).
Bibliography 277
R.A. Kruger, P. Liu, “Photoacoustic ultrasound: pulse production and detection in 0.5-percent
liposyn,” Med. Phys. 21, 1179–1184 (1994).
R.A. Kruger, P. Liu, Y. Fang, C. Appledorn, “Photoacoustic ultrasound (PAUS)-reconstruction
tomography,” Med. Phys. 22, 1605–1609 (1995).
C. Kuhl, “The current status of breast MR imaging. Part I. Choice of technique, image inter-
pretation, diagnostic accuracy, and transfer to clinical practice,” Radiology 244, 356–
378 (2007).
M.L.J. Landsman, G. Kwant, G.A. Mook, W.G. Zijlstra, “Light-absorbing properties, stability,
and spectral stabilization of indocyanine green,” J. Appl. Physiol. 40, 575–583 (1976).
S.B. Lee, M. Hassan, R. Fisher, O. Chertov, V. Chernomordik, G. Kramer-Marek G, et al.
“Affibody molecules for in vivo characterization of HER2-positive tumors by near-
infrared imaging,” Clin. Cancer Res. 14, 3840–3849 (2008).
Y. Li, N. Wood, D. Yellowlees, P.K. Donnelly. Anticancer Res. 19, 1223–1228 (1999).
C. Li, H. Zhao, B. Anderson, H. Jiang, “Multispectral breast imaging using a ten-wavelength,
64 x 64 source/detector channels silicon photodiode-based diffuse optical tomography
system,” Med. Phys. 33 (3), 627–636 (2006).
M. Lipowska, G. Patonay, L. Strekowski. Synth. Commum. 23, 3087–3094 (1993).
E. Mace, G. Montaldo, I. Cohen, M. Baulac, M. Fink, M. Tanter, “Functional ultrasound imag-
ing of the brain,” Nature Methods 8, 662–685 (2011).
S. Madsen, M. Patterson, B. Wilson, “The use of india ink as an optical absorber in tissue-
simulating phantoms,” Phys. Med. Biol. 37, 985–993 (1992).
H. Maeda,Y. Matsumura, “Tumoritropic and lymphotropic principles of macromolecular drugs,”
Crit Rev Ther Drug Carrier Syst 6 (3), 193–210 (1989).
P.S. Mangan, C. Scott, J. Williamson, E. Bertram, “Aberrant neuronal physiology in the basal
nucleus of the amygdala in a model of chronic limbic epilepsy,” Neuroscience 101 (2),
377–391 (2000).
S. Manohar, A. Kharine, J. C. G. van Hespen, W. Steenbergen, T. G. van Leeuwen, J. Biomed.
Opt. 9, 1172 (2004).
S. Manohar, S.E. Vaartjes, J.C.G. van Hespen, J.M. Klaase, F.M. van den Engh, W. Steenbergen,
T.G. van Leeuwen, “Initial results of in vivo noninvasive cancer imaging in the human
breast using near-infrared photoacoustics,” Opt. Express 15, 12277–12285 (2007).
J. Mansell et al., “Bone, not cartilage, should be the major focus in osteoarthritis,” Nature
Clinical Practice Rheumatology 3, 306–307 (2007).
T.F. Massoud, S.S. Gambhir. Genes. Dev. 17, 545–580 (2003).
R. Matloub, A. Hadad, A. Mazzalai, N. Chidambaram, G. Moulard, Appl. Phys. Lett. 102,
152903 (2013).
F. Meric, N.Q. Mirza, G. Vlastos, T.A. Buchholz, H.M. Kuerer, G.V. Babiera, S.E. Singletary,
M.I. Ross, F.C. Ames, B.W. Feig, S. Krishnamurth, G.H. Perkins, M.D. McNeese, E.A.
Strom, V. Valero, K.K. Hunt, “Positive surgical margins and ipsilateral breast tumor
recurrence predict disease-specific survival after breast-conserving therapy,” Cancer 97
(4), 926–933 (2003).
A. Moore, E. Marecos, A. Bogdanov Jr., R. Weissleder, Radiology 214, 568–574 (2000).
P. Muralt, IEEE Trans. Ultrason. Ferroelectr. Freq. Control 47, 903–915 (2000).
P. Muralt, N. Ledermann, J. Baborowski, A. Barzegar, S. Gentil, B. Belgacem, S. Petitgrand,
A. Bosseboeuf, N. Setter, IEEE Trans. Ultrason. Ferroelectr. Freq. Control 52 (12),
2276–2288 (2005).
S.L. Myers, K. Dines, D. Brandt, K. Brandt, M. Albrecht, “Experimental assessment by high
frequency ultrasound of articular cartilage thickness and osteoarthritis changes,” J.
Rheumatol. 22, 109–116 (1995).
M. Nakazawa, M, Tararu, K. Nakamura, S. Ueha, A. Maezawa, Jpn. J. Appl. Phys. 46 (7B),
4466 (2007).
J.J. Niederhauser, M. Jaeger, M. Frenz, App. Phys. Lett. 85 (5), 846 (2004).
278 Bibliography
S.J. Norton, T. Vo-Dinh, “Optoacoustic diffraction tomography: analysis of algorithms,” J.

Opt. Soc. Am. A, 20, 1859–1866 (2003).
A.A. Oraevsky, S.L. Jacques, R.O. Esnaliev, F.K. Tittel, “Laser-based optoacoustic imaging in
biological tissues,” Proc. SPIE (2124A), 122–128 (1994).
M. Ostergaard, M. Stoltenberg, P. Lovgreen-Nielsen, B. Volck, C. Jensen, I. Lorenzen, “Magnetic
resonance imaging determined synovial membrane and joint effusion volumes in rheu-
matoid arthritis and osteoarthritis,” Arthritis and Rheumatism 40, 1856–1867 (1997).
S. Pahl, “Optical properties spectra,” 2003, http://omlc.ogi.edu/spectra/index.html.
H.S. Pan, S. Lee, L. Huang, Y. Chen, “Combined positron emission tomography and tumor
markers for detecting recurrent ovarian cancer,” Archives of Gynecology and Obstetrics
283 (2), 335–341 (2011).
H.E. Panula, M. Hyttinen, J. Arokoski, T. Langsjo, A. Pelttari, I. Kivitanta, H. Helminen,
“Articular cartilage superficial zone collagen birefringence reduced and cartilage thick-
ness increased before surface fibrillation in experimental osteoarthritis,” Ann. Rheum.
Dis. 57, 237–245 (1998)S.C. Partridge, J.E. Gibbs, Y. Lu, L.J. Esserman, D. Tripathy,
D.S. Wolverton, H.S. Rugo, E.S. Hwang, C.A. Ewing, N.M. Hylton, “MRI measure-
ments of breast tumor volume predict response to neoadjuvant chemotherapy and recur-
rence-free survival,” Am J Roentgenol 184 1774–1781 (2005).
K. D. Paulsen, H. Jiang, “Enhanced frequency-domain optical image reconstruction in tissues
through total-variation minimization,” App. Opt. 35, 3447–3458 (1996).
G.W. Paxinos, The rat brain. Academic Press, 1998.
P. Peloschek, G. Langs, M. Weber, J. Sailer, M. Reisegger, H. Imhof, H. Bischol, F. Kainberger,
“An automatic model-based system for joint space measurements on hand radiographs:
Initial experience,” Radiology 243, 855–862 (2007).
R. Philip, A. Penzkofer, W. Baumler, R. M. Szeimies, C. Abels, “Absorption and fluorescence
spectroscopic investigation of indocyanine,” J. Photochem. Photobiol. A: Chemistry 96,
137–148 (1996).
B. Pogue, M. Patterson, “Frequency-domain optical absorption spectroscopy of finite tissue
volumes using diffusion theory,” Phys. Med. Biol. 39, 1157–1180 (1994).
V. Prapavat, W. Runge, J. Mans, A. Krause, J. Beuthan, G. Muller, “The development of a fin-
ger joint phantom for the optical simulation of early inflammatory rheumatic changes,”
Biomedizinische Technik 42, 319–326 (1997).
W. Press, B. Flannery, S. Teukolsky, W. Vetterling, Numerical Recipes, Cambridge Univ.
Press (1992).
D. Psaltis, S. R. Quake, C. Yang, Nature 442 (7101), 381 (2006).
N. Qatanani, A. Barham, and Q. Heeh, “Existence and uniqueness of the solution of the coupled
conduction radiation energy transfer on diffuse gray surfaces,” Surveys in Mathematics
and Its Applications, 2, 43– 58 (2007).
K. Ren, G. Bal, A.H. Hielscher, “Transport- and diffusion-based optical tomography in small
domains: a comparative study,” Appl. Opt. 46, 6669–6679 (2007).
P.M. Rodgers, J. Ward, C.J. Baudouin, J.P. Ridgway, P.J. Robinson, “Dynamic contrast-
enhanced MR imaging of the portal venous system: comparison with x-ray angiogra-
phy,” Radiology 191, 741–774 (1994).
C. Rousseau, A. Devillers, C. Sagan, L. Ferrer, B. Bridji, L. Campion, M. Ricaud, E.
Bourbouloux, I. Doutriaux, M. Clouet, D. Berton-Rigaud, C. Bouriel, V. Delecroix, E.
Garin, S. Rouquette, I. Resche, P. Kerbrat, JF. Chatal, M. Campone, “Monitoring of
early response to neoadjuvant chemotherapy in stage II and III breast cancer by [18F]
fluorodeoxyglucose positron emission tomography,” J Clin Oncol 24 5366–5372 (2006).
L.I. Rudin, S. Osher, E. Fatemi, “Nonlinear total variation based noise removal algorithm,”
Physica D 60, 259–268 (1992).
M. Sabuncu, Entropy-based image registration, The Department of Electrical Engineering,
Princeton University, pp. 152 (2006).
Bibliography 279
M. Satpathy, L. Wang, R. Zielinski, W. Qian, M. Lipowska, J. Capala, G.Y. Lee, H. Xu, A.

Wang, H. Mao, L. Yang, Active targeting using HER-2-affiboday-conjugated nanopar-
ticles enabled sensitive and specific imaging of orthotopic HER-2 positive ovarian
tumors. Small DOI: 10.1002/smll.201301593 (2013).
J.A. Schaar, J.E. Muller, E. Falk, et al. “Terminology for high-risk and vulnerable coronary
artery plaques,” Report of a meeting on the vulnerable plaque, June 17 and 18, 2003,
Santorini, Greece. Eur Heart J 25, 1077–1082 (2004).
A.P. Schroeder, E. Falk, “Vulnerable and dangerous coronary plaques,” Atherosclerosis 118
Suppl: 141–149 (1995).
A. Shaaban, M. Rezvani, “Ovarian cancer: detection and radiologic staging,” Clinical
Obstetrics and Gynecology 52 (1), 73–93 (2009).
J. Shah, L. Ma, K. Sokolov, K. Johnston, T. Milner, S.Y. Emelianov, S. Park, S. Aglyamov, T.
Larson, “Photoacoustic imaging and temperature measurement for photothermal cancer
therapy,” J Biomed Opt 13, 034024 (2008).
M. Sharan, E. Vovenko, A. Vadapalli, A. Popel, R. Pittman, “Experimental and theoretical
studies of oxygen gradients in rat pial micro vessels,” J. Cerebr. Blood F. Met. 28, 1597–
1604 (2008).
P. Sharples, A. Stuart, D. Matthews, A. Aynsley-Green, J. Eyre, “Cerebral blood flow and
metabolism in children with severe head injury. Part 1: Relation to age, Glasgow coma
score, outcome, intracranial pressure, and time after injury,” J. Neurol. Neurosurg.
Psychiatry 58, 145–152 (1995).
S. Shelton, M. Chen, H. Park, B. Boser, I. Izyumin, R. Przybyla, T. Frey, M. Judy, K. Nunan,
F. Sammoura, IEEE Ultrasonics Symposium (IUS), Rome, (2009), pp. 402–405.
B. Sigal-Zafrani, J.S. Lewis, K.B. Clough, A. Vincent-Salomon, A. Fourquet, M. Meunier,
M.C. Falcou, X. Sastre-Guarau, “Histological margin assessment for breast ductal car-
cinoma in situ: precision and implications,” Mod Pathol. 17 (1), 81–88 (2004).
D. Sipkins, X.B. Wei, J.W. Wu, J.M. Runnels, D. Cote, T.K. Means, A.D. Luster, D.T. Scadden,
C.P. Lin, “in vivo imaging of specialized bone marrow endothelial microdomains for
tumour engraftment,” Nature 435, 969–973 (2005).
I.M. Smith, D.V. Griffiths, Programming the finite element method, John Wiley & Sons,
Chichester, UK (2004).
D.K. Sokol, O. Markand, E. Daly, T. Luerssen, M. Malkoff, “Near infrared spectroscopy
(NIRS) distinguishes seizure types,” Seizure 9, 323–327 (2000).
C. Song, N.-T. Nguyen, S.-H. Tan, A.K. Asundi, “Modelling and optimization of micro opto-
fluidic lenses,” Lab on a Chip 9, 1178–1184 (2009).
C. Song, N.-T. Nguyen, S.-H. Tan, A. Asundi, “A tuneable micro-optofluidic biconvex lens
with mathematically predictable focal length,” Microfluidics and Nanofluidics 9, 889–
896 (2010).
C. Song, N.-T. Nguyen, Y. Yap, T.-D. Luong, and A. Asundi, “Multi-functional, optofluidic,
in-plane, bi-concave lens: tuning light beam from focused to divergent,” Microfluidics
and Nanofluidics 10, 671–678 (2011).
C. Song, L. Xi, H. Jiang, “Liquid acoustic lens for photoacoustic tomography,” Opt. Lett. 38,
2930–2933 (2013).
S. Srinivasan, B.W. Pogue, S. Jiang et al., “Interpreting hemoglobin and water concentration,
oxygen saturation, and scattering measured in vivo by near-infrared breast tomography,”
Proc Natl Acad Sci USA 100, 12349–12354 (2003).
B.J. Steinhoff, G. Herrendrof, C. Kurth, “Ictal near infrared spectroscopy in temporal lobe
epilepsy: a pilot study,” Seizure 5, 97–101 (1996).
N. Sugiura, S. Morita, Applied Optics 32, 4181–4186 (1993).
A. Taghian, M. Mohiuddin, R. Jagsi, S. Goldberg, E. Ceilley, S. Powell, “Current perceptions
regarding surgical margin status after breast-conserving therapy: results of a survey,”
Ann Surg. 241 (4), 629–639 (2005).
280 Bibliography
A.C. Tam, “Applications of photoacoustic sensing techniques,” Rev. Mod. Phys. 58, 381–
431 (1986).
J. Tammela, S. Lele, “New modalities in detection of recurrent ovarian cancer,” Current
Opinion in Obstetrics and Gynecology 16 (1), 5–9 (2004).
H.Y. Tan, W.K. Loke, N.T. Nguyen, Sensors and Actuators B: Chemical 151 (1), 133 (2010).
T. Tarvainen, M. Vauhkonen, V. Kolehmainen, J. P. Kaipio, “Hybrid radiative transfer-diffu-
sion model for optical tomography,” Appl. Opt. 44, 876–886 (2005).
L. Terslev, S. Torp-Pedersen, E. Qvistgaard, P. von der Recke, H. Bliddal, “Doppler ultrasound
findings in healthy wrists and finger joints,” Ann Rheum Dis 63, 644–648 (2004).
J. Tervo, P. Kolmonen, M. Vauhkonen, L. Heikkinen, J. Kaipio, “Data fitting model for the ker-
nel of integral operator from radiation therapy,” Inverse Probl. 15, 1345–1361 (1999).
M. Thom, G. Mathern, J. Cross, E. Bertram, “Mesial temporal lobe epilepsy: How do we
improve surgical outcome?” Ann Neurol. 68 (4), 424–434 (2010).
A. Tsai, P. Johnson, M. Intaglietta, “Oxygen gradients in the microcirculation,” Physiol. Rev.
83, 933–963 (2003).
K. Toutouzas, S. Vaina, M. Riga, C. Stefanadis, “Evaluation of dissection after coronary
stent implantation by intravascular optical coherence tomography,” Clin Cardiol 32,
E47–48 (2009).
J.A. Tyler, P. Watson, H. Koh, N. Herrod, M. Robson, L. Hall, “Detection and monitoring of
progressive degeneration of osteoarthritic cartilage by MRI”, Acta Orthop. Scand. 66
Suppl. 266, 130–138 (1995).
S. Vaithilingam, I.O. Wygant, P.S. Kuo, X. Zhuang, Ö Oralkan, P.D. Olcott, B.T. Khuri-Yakub,
Proc. SPIE, 6086, 608603(2006).
P. M. van den Berg, R. E. Kleinmann, “A total variation enhanced modified gradient algorithm
for profile reconstruction,” Inverse Probl. 11, L5–L10 (1995).
H. van Staveren, C. Moes, J. van Marle, A. Prahl, M. van Germert, “Light scattering in
Intralipid-10% in the wavelength range of 400–1100nm,” Appl. Opt. 30, 4507–4514 (1991).
R. Virmani, F. Kolodgie, A. Burke, et al., “Lessons from sudden coronary death: a compre-
hensive morphological classification scheme for atherosclerotic lesions,” Arterioscler
Thromb Vasc Biol 20, 1262–1275 (2000).
C.R. Vogel, and M.E. Oman, “Iterative methods for total variation denoising,” SIAM J. Sci.
Comput. 17, 227–238 (1996).
G.K. von Schulthess, H.C. Steinert, T.F. Hany, “Integrated PET/CT: current applications and
future directions,” Radiology 238, 405–422 (2006).
Z. Wang, A. Bovik, “A universal image quality index,” IEEE Signal Process. Lett. 9, 81–84 (2002).
X. Wang, D.L. Chamberland, D.A. Jamadar, “Noninvasive photoacoustic tomography of
human peripheral joints toward diagnosis of inflammatory arthritis,” Opt. Lett. 32,
3002–3004 (2007).
B. Wang, L. Xiang, M.S. Jiang, J. Yang, Q. Zhang, P.R. Carney, H. Jiang, Biomed. Opt. Express
3 (6), 1427 (2012).
E. Warner, D.B. Plewes, K.A. Hill et al., “Surveillance of BRCA1 and BRCA2 mutation car-
riers with magnetic resonance imaging, ultrasound, mammography, and clinical breast
examination,” JAMA 292 (11), 1317–1325 (2004).
E. Watanabe, A. Maki, F. Kawaguchi, Y. Yamaguchi, H. Koizumi, Y. Mayanagi, “Noninvasive
cerebral blood volume measurement during seizures using multichannel near infrared
spectroscopic topography,” J Biomed. Opt. 5, 287–290 (2000).
G.E. Wnek, G.L. Bowlin, Encyclopedia of biomaterials and biomedical engineering, Marcel
Dekker, New York, (2004).
D.W. Wu, Q.F. Zhou, K.K. Shung, C.G. Liu, F.T. Djuth, IEEE Ultrasonics Symposium (IUS),
Beijing, (2008), pp. 1222–1225.
Bibliography 281
L. Wu, H. Xie, “A large vertical displacement electrothermal bimorph microactuator with very
small lateral shift,” Sens and Actuators A Phys. 145–146, 371–379 (2008).
L. Xi, X. Li, L. Yao, H. Jiang, Med. Phys. 39, 2584 (2012).
L. Xi, C. Duan, H. Xie, H. Jiang, “Miniature probe combining optical-resolution photoacous-
tic microscopy and optical coherence tomography for in vivo microcirculation study,”
Appl. Opt. 52, 1928–1931 (2013).
W. Xia, D. Piras, M. Heijblom, J. Hespen, S. Veldhoven, C. Prins, W. Steenbergen, T. Leeuwen,
S. Manohar, Proc. of SPIE-OSA 8090, 80900L (2011).
L. Xiang, B. Wang, L. Ji, H. Jiang, “4D photoacoustic tomography,” Nature Scientific Reports
3, 1113 (2013).
H. Xie, Y. Pan, G.K. Fedder, “Endoscopic optical coherence tomographic imaging with a
CMOS MEMS micromirror,” Sensors & Actuators: A, 103, 237–241 (2003).
Y. Xu, X. Gu, L.L. Fajardo, H. Jiang, “In vivo breast imaging with diffusion optical tomogra-
phy based on higher-order diffusion equations,” Appl. Opt. 42, 3163–3169 (2003).
L. Yang, X.H. Peng, Y.A. Wang, X. Wang, Z. Cao, C. Ni, P. Karna, X. Zhang, W.C. Wood, X.
Gao, S. Nie, H. Mao. Clin. Cancer Res. 15, 4722–4732 (2009).
L. Yao, H. Jiang, “Finite-element-based photoacoustic tomography in time-domain,” J. Opt. A:
Pure Appl. Opt. 11, 085301 (2009).
L. Yao, H. Jiang, “Enhanced photoacoustic tomography using total variation minimization,”
Applied Optics 50, 5031–5041 (2011).
L. Yao, Y. Sun, H. Jiang, “Quantitative photoacoustic tomography based the radiative transfer
equation,” Opt. Lett. 34, 1765–1767(2009).
Y. Honda, P. Fitzgerald, “Frontiers in intravascular imaging technologies,” Circulation 117,
2024–2037 (2008).
L. Yin, Q. Wang, Q. Zhang, H. Jiang, “Tomographic imaging of absolute optical absorption
coefficient in turbid media using combined photoacoustic and diffusing light measure-
ments”, Opt. Lett. 32, 2556–2558 (2007).
P. Yock, E. Johnson, D. Linker, “Intravascular ultrasound: developmentband clinical poten-
tial,” Am J Card Imaging 2, 185–193 (1988).
Z. Yuan, H. Jiang, “Quantitative photoacoustic tomography: recovery of optical absorption
coefficient of heterogeneous media,” Appl. Phys. Lett. 88, 213301-1-3 (2006).
Z. Yuan, H. Jiang, “An image reconstruction scheme that combines modified Newton method
and efficient initial guess estimation for optical tomography of finger joints,” Appl. Opt.
46, 2757–2768 (2007).
Z. Yuan, H. Jiang, “A calibration-free, one-step method for quantitative photoacoustic tomog-
raphy,” Med. Phys. 39, 6895–6899 (2012).
Z. Yuan, C. Wu, H. Zhao, H. Jiang, “Imaging of small nanoparticle-containing objects by finite
element-based photoacoustic tomography,” Opt. Lett. 30, 3054–3056 (2005).
Z. Yuan, Q. Zhang, H. Jiang, “Simultaneous reconstruction of acoustic and optical properties
of heterogeneous media by quantitative photoacoustic tomography,” Opt. Express 14,
6749–6754 (2006).
Z. Yuan, Q. Wang, H. Jiang, “Reconstruction of optical absorption coefficient maps of het-
erogeneous media by photoacoustic tomography coupled with diffusion equation based
regularized Newton method,” Opt. Express 15, 18076–18081 (2007).
Z. Yuan, Q. Zhang, E.S. Sobel, H. Jiang, “Tomographic x-ray guided three-dimensional diffuse
optical tomography of osteoarthritis in the finger joints,” J. Biomed. Opt. 13, 044006-
1-10 (2008).
Z. Yuan, X. Hu, H. Jiang, “A higher order diffusion model for three dimensional photon migra-
tion and image reconstruction in optical tomography,” Phys Med Biol. 54, 65–88 (2009).
Y. Yuan, S. Yang, D. Xing, Applied Physics Letters 100, 023702-3 (2012)
282 Bibliography
J.M. Zara, S. Yazdanfar, K.D. Rao, J.A. Izatt, S.W. Smith, “Electrostatic micromachine scan-
ning mirror for optical coherence tomography,” Optics Letters 22, 628–630 (2003).
A.D.L. Zerda, C. Zavaleta, S. Keren, S. Vaithilingam, S. Bodapati, Z. Liu, J. Levi, B.R. Smith,
T.J. Ma, O. Oralkan, Z. Cheng, X. Chen, H. Dai, B.T. Khuri-Yakub and S.S. Gambhir,
Nat. Nanotech. 3, 557–562 (2008).
Q. Zhang, P. A. Lewin, Acoustics 20, 77 (1995).
D.Y. Zhang, N. Justis, Y. . Lo, App. Phys. Lett. 84 (21), 4194 (2004).
V.P. Zharov, E.I. Galanzha, E.V. Shashkov, N.G. Khlebtsov, V.V. Tuchin, “In vivo photoacous-
tic flow cytometry for monitoring of circulating single cancer cells and contrast agents,”
Opt. Lett. 31, 3623 (2006).
J. Zhou, S.S. Talathi, A. Cadotte, Z. Liu, G.L. Holmes, P.R. Carney, “The Effects of
Hippocampus Ca1 Single Neuron Firing Properties on Interictal Spike Patterns dur-
ing Seizure Onset in a Rat Model of Temporal Lobe Epilepsy.” Epilepsia 50, 317–
317 (2009).
Electrical Engineering
Photoacoustic Tomography
"… a very comprehensive collection of photoacoustic imaging methods, covering
everything from the basics of photoacoustic generation and image reconstruction
methods to the various implementations, including numerous applications. This is
the first time the topic of photoacoustic imaging has been covered in this detail
by a single author."
—Günther Paltauf, Karl-Franzens-Universität Graz, Austria
The concept of photoacoustic tomography (PAT) emerged in the mid-1990s,
and the field of PAT is now rapidly moving forward. Presenting the research of
a well-respected pioneer and leading expert, Photoacoustic Tomography
is a first-of-its-kind book covering the underlying principles and practical
applications of PAT in a systematic manner. Written in a tutorial format,
the text:
• Addresses the fundamentals of PAT, the theory on photoacoustic

effect, image reconstruction methods, and instrumentation
• Details advanced methods for quantitative PAT, which allow the recovery
of tissue optical absorption coefficient and/or acoustic properties
• Explores the development of several image-enhancing schemes,
including both software and hardware approaches
• Examines array-based PAT systems that are the foundation for the
realization of 2-D, 3-D, and 4-D PAT
• Discusses photoacoustic microscopy (PAM) and combinations
of PAT/PAM with other imaging methods
• Considers contrast agents–based molecular PAT, with both
nontargeted and cell receptor–targeted methods
• Describes clinical applications and animal studies in breast cancer
detection, osteoarthritis diagnosis, seizure localization, intravascular
imaging, and image-guided cancer therapy
Photoacoustic Tomography is an essential reference for graduate

students, researchers, industry professionals, and those who wish to
enter this exciting field.
K24289

Tomografia Fotoacustica 2015!! PDF

Загружено:

Сведения о документе

Оригинальное название

Авторское право

Доступные форматы

Поделиться этим документом

Поделиться или встроить документ

Параметры публикации

Этот документ был вам полезен?

Это неприемлемый материал?

Авторское право:

Доступные форматы

Tomografia Fotoacustica 2015!! PDF

Загружено:

Авторское право:

Доступные форматы

Aperture

Boca Raton London New York

CRC Press is an imprint of the

No claim to original U.S. Government works

International Standard Book Number-13: 978-1-4822-6104-2 (eBook - PDF)

Chapter 1 Fundamentals of Photoacoustic Tomography....................................... 1

Chapter 2 Quantitative Photoacoustic Tomography............................................ 15

Chapter 3 Image Enhancement Software and Hardware Approaches................. 47

3.4.3.1 Comparison of the Transport and

Chapter 4 Transducer Array-Based Photoacoustic Tomography: 2D, 3D,

Chapter 5 Photoacoustic Microscopy and Photoacoustic Computed

5.3.2.1 Phantom Experiments................................ 137

Chapter 6 Multimodal Approaches.................................................................... 169

Chapter 7 Contrast Agents–Based Molecular Photoacoustic Tomography....... 193

7.3.1.2 Preparation of NIR830-ATF-IONP

Chapter 8 Clinical Applications and Animal Studies........................................ 223

8.1.3 3D Single-Spectral Quantitative PAT....................... 233

Photoacoustic tomography (PAT), or optoacoustic tomography (OAT), is based

Φ = Φ (x, y, 0) = E/A (1.1)

The radiant exposure for a large depth z is estimated by,

dV/V = −κp + βΔT (1.6)

where κ denotes the isothermal compressibility, β is the thermal coefficient of vol-

If we define a set of coefficients as Grueneisen parameter, Γ (Gusev and

Γ = β/(κρCV ) = βvs2 /C p (1.9)

Equation (1.8) is rewritten in consideration of Equations (1.4), (1.5), and (1.9),

Γ (T0 ) = 0.0043 + 0.0053T0 (1.11)

in which T0 is the temperature in degrees Celsius. p 0 is often specified as the initial

τ p < d /vs (1.12)

τth < d 2 /α th (1.13)

in which α th is the thermal diffusivity (m2/s) and dc ⊕1/ eff .

Φ0 = E/πr2= 30 mJ/3.14 cm2 = 9.55 mJ/cm2

Assuming μa = 25 cm –1, the absorbed light energy density is calculated as,

Ψ = 25 × 9.55 = 238.75 mJ/cm3 = 2.38 × 105 J/m3

T ( x , y , z ) = Ψ( x , y , z )/(ρC p ) = 238.75 mJ/cm3 /(1g/cm3 × 4 Jg −1K −1) = 59.7 mk

If a homogeneous elastic medium is assumed, Equation (1.17) is further written,

Equation (1.18) is the general PA wave equation (without acoustic attenuation) in

where P is the pressure wave in frequency-domain; k = ω/vs is the wave number

O = vo2 /vs2 − 1 + iAv0 /k0 vs2 (1.24)

where v0 is the constant acoustic velocity in a reference/coupling medium, and A is

1.2 IMAGE RECONSTRUCTION METHODS

1.2.2 Iterative Nonlinear Algorithm

1.2.2.1 Frequency-Domain FE-Based Algorithm

Expanding acoustic pressure, P as the sum of coefficients multiplied by a set of

= − ik0 v0βΨ/CP ψ i (1.27)

[ A]{P } = {B} (1.29)

{P} = {P1 , P2 ,…, PN }T (1.32)

where Ψ R = Ψ  R − Ψ R , and Ψ I = Ψ  I − Ψ I . If the assumed optical property dis-

∂ P1 /∂Ψ R ,1  ∂ P1 /∂Ψ R , K ∂ P1 /∂Ψ I ,1  ∂ P1 /∂Ψ I , L

χ = ( Ψ R ,1 , Ψ R ,2 ,…, Ψ R ,K , Ψ I ,1 , Ψ I ,2 ,…, Ψ I , L )T (1.36)

P o = ( P1o , P2o ,…, PMo )T (1.37)

P c = ( P1c , P2c ,…, PMc )T (1.38)

1.2.2.2 Time-Domain FE-Based Algorithm

where I (t ) = δ(t − t0 ) is typically assumed.

Expanding p as the sum of coefficients multiplied by a set of basis function ψj,

The Bayliss-Turkel radiation boundary conditions are employed here (Jin

where Bm is a sequence of linear differential operators,

and Bm is subject to the accuracy,

This method provides an even more accurate approximation to the Sommerfeld

The boundary term based on Equation (1.42) is rewritten

[ K ]{p} + [C ]{p} + [ M ]{

where the elements of the matrix are written

{p} = {p1 , p2 ,…, pN }T (1.52)

{P}t + t = {P}t + [(1 − δ){P}t + δ{P}t + t ] t (1.55)

1.2 IMAGE RECONSTRUCTION METHODS

1.2.2 Iterative Nonlinear Algorithm

where Ψ R = Ψ R − Ψ R , and Ψ I = Ψ I − Ψ I . If the assumed optical property dis-

∂ P1 /∂Ψ R ,1 ∂ P1 /∂Ψ R , K ∂ P1 /∂Ψ I ,1 ∂ P1 /∂Ψ I , L

[ K ]{p} + [C ]{p} + [ M ]{

{P}t + t = {P}t + [(1 − δ){P}t + δ{P}t + t ] t (1.55)

{P}t + t = {P}t + t (1 − δ){P}t + δ t{P}t + t (1.59)

1.2.3 A-Line/B-Mode Image Formation Method

2.1 RECOVERY OF OPTICAL ABSORPTION

2.2 RECOVERY OF OPTICAL ABSORPTION

2.3 RECOVERY OF OPTICAL ABSORPTION

2.4 RECOVERY OF OPTICAL ABSORPTION

2.5 SIMULTANEOUS RECOVERY OF OPTICAL ABSORPTION

The Jacobian matrix, J, is denoted as J = [ J O ,λ ,ω , J ci ,λ ,ω ], where J O ,λ ,ω and J ci ,λ ,ω

3.1 DUAL MESH SCHEME

3.2 ADJOINT SENSITIVITY METHOD

3.3.2 Examples: Reduction of Noise Effect