412

IEEE TRANSACTIONS ON BIOMEDICAL CIRCUITS AND SYSTEMS, VOL. 9, NO. 3, JUNE 2015

A Thorax Simulator for Complex Dynamic

Bioimpedance Measurements With Textile Electrodes
Mark Ulbrich, Jens Mhlsteff, Daniel Teichmann, Student Member, IEEE,
Steffen Leonhardt, Senior Member, IEEE, and Marian Walter, Senior Member, IEEE

AbstractBioimpedance measurements on the human thorax

are suitable for assessment of body composition or hemodynamic
parameters, such as stroke volume; they are non-invasive, easy in
application and inexpensive. When targeting personal healthcare
scenarios, the technology can be integrated into textiles to increase
ease, comfort and coverage of measurements. Bioimpedance is
generally measured using two electrodes injecting low alternating
currents (0.510 mA) and two additional electrodes to measure
the corresponding voltage drop. The impedance is measured either spectroscopically (bioimpedance spectroscopy, BIS) between
5 kHz and 1 MHz or continuously at a xed frequency around
100 kHz (impedance cardiography, ICG). A thorax simulator
is being developed for testing and calibration of bioimpedance
devices and other new developments. For the rst time, it is
possible to mimic the complete time-variant properties of the
thorax during an impedance measurement. This includes the
dynamic real part and dynamic imaginary part of the impedance
and an adjustable base
with a peak-to-peak value of 0.2
. Another novelty is adimpedance
justable complex electrode-skin contact impedances for up to
8 electrodes to evaluate bioimpedance devices in combination
with textile electrodes. In addition, an electrocardiographic signal
is provided for cardiographic measurements which is used in
ICG devices. This provides the possibility to generate physiologic impedance changes, and in combination with an ECG, all
parameters of interest such as stroke volume (SV), pre-ejection
can be simulated.
period (PEP) or extracellular resistance
The speed of all dynamic signals can be altered. The simulator
was successfully tested with commercially available BIS and ICG
devices and the preset signals are measured with high correlation
.

it may be difcult to maintain adequate medical care. Therefore, health monitoring technology will enter the personal
environment of humans. Among the available technologies,
bioimpedance is a potential candidate, as it facilitates assessment of body composition and cardiac status in an easy-to-use,
inexpensive and non-invasive way. Once integrated into textiles, the monitoring of patients 24/7 becomes possible. The
integration of bioimpedance devices into textiles is challenging
since textile electrodes imply a high load for current sources
and often the impedance may be larger than the value given in
the specications.
In this work, a simulator was developed for validating bioimpedance instruments which aim to assess the
bioimpedance on the human thorax. Since the simulator produces a time-variant complex impedance, both bioimpedance
spectroscopy (BIS) devices and impedance cardiography
(ICG) devices (or combinations thereof) can be tested. An
electrocardiographic signal (ECG) is produced in parallel
to the impedance for ICG measurements since ICG devices
calculate certain parameters from the combination of ECG and
impedance measurements. In addition, electrode-skin contact
impedances caused by textile electrodes can be altered to test
current sources of different devices. Previous studies concentrated on resistive simulations, thereby neglecting all capacitive
effects and focusing only on standard ICG devices [2][4].
Below we describe the measurement methods, the technical
implementation of the simulator, and the measurement results.

Index TermsBioimpedance, cardiography, impedance, simulator, spectroscopy, textile electrodes.

II. BIOIMPEDANCE MEASUREMENTS

I. INTRODUCTION

EMOGRAPHIC changes are leading to an increasingly

aging society with additional burden on medical resources and the welfare system [1]. Especially in rural areas
Manuscript received January 29, 2014; revised April 17, 2014 and May 21,
2014; accepted July 05, 2014. Date of publication August 19, 2014; date of
current version May 22, 2015. This work was supported by HeartCycle, an EU
project focusing on compliance and effectiveness in heart failure and coronary
heart disease closed-loop management, and by Philips Research Europe. This
paper was recommended by Associate Editor E. Jovanov.
M. Ulbrich, D. Teichmann, S. Leonhardt, and M. Walter are with the Philips
Chair for Medical Information Technology, RWTH Aachen University, 52074
Aachen, Germany (e-mail: ulbrich@hia.rwth-aachen.de).
J. Mhlsteff is with Philips Research, High Tech Campus 34, 5656 AE Eindhoven, The Netherlands.
Color versions of one or more of the gures in this paper are available online
at http://ieeexplore.ieee.org.
Digital Object Identier 10.1109/TBCAS.2014.2337372

Bioimpedance measurements do not aim at the acquisition

of the body's biopotentials, but at assessment of the complex
impedance of the body evoked by an injected alternating current. Due to safety regulations, the current is limited to 0.1 mA
below 1 kHz and to 10 mA above 100 kHz [5]. Organic tissues
show three frequency dependent dispersion ranges between
1 Hz and 1 GHz: -dispersion (1 Hz10 kHz), -dispersion
(10 kHz10 MHz) and -dispersion (10 MHz1 GHz) [6].
Since pathologic and pathophysiologic processes lead to the
highest changes of the impedance between 5 kHz and 1 MHz,
bioimpedance is measured mainly in the -dispersion range.
Here, the Maxwell-Wagner effect is dominant, which is based
on polarisation or relaxation of the polarisation at boundary
surfaces such as the cellular membrane [7].
For low frequencies, the current ows around the tissues cells
through the extracellular water (ECW) since the cell membrane
acts as a capacitance. For high frequencies, ECW and intracellular water (ICW) are penetrated by the current (Fig. 1). The

1932-4545 2014 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

ULBRICH et al.: A THORAX SIMULATOR FOR COMPLEX DYNAMIC BIOIMPEDANCE MEASUREMENTS

413

Fig. 1. Current ow and electrical equivalent circuit: extracellular water

(ECW) intracellular water (ICW).

electrical equivalent circuit of a tissue is composed of a resistor

representing the extracellular space
in parallel to a resistor
representing the intracellular space
in series with a capacitor
for the isolating properties of the cellular membrane.
Interestingly, this Cole model of tissues is basically also valid
for measurements on body parts or the whole body [7].
A. Bioimpedance Spectroscopy (BIS)
BIS describes spectral bioimpedance measurements between
5 kHz and 1 MHz. Classically, bioimpedance measurements are
performed using two outer electrodes to inject the current
and two inner electrodes for the voltage assessment
.
Thus, the complex impedance can be calculated by Ohm's law
(1)
Using the Cole model, the frequency dependent impedance is
described as follows:

Fig. 2. Thoracic impedance

characteristic points.

and ICG wave

with its

B. Impedance Cardiography (ICG)

ICG is a bioimpedance measurement technique which aims
at assessing cardiac health status by continuous measurement of
the absolute impedance at one xed frequency around 100 kHz.
This allows time-dependent hemodynamic parameters to be extracted from the impedance curve using its temporal derivative
and certain characteristic points (Fig. 2).
The C-point is simply dened by the maximum of the ICG
curve. The B and X-points can be extracted in several ways,
e.g., by using the zero crossing of the ICG wave before the
C-point for B-point extraction and the global minimum after the
C-point for X-point extraction [8]. To estimate stroke volume
(SV) using these characteristic points, several model assumptions have been made by different groups. A very common one
is the Bernstein-Sramek equation:
(6)

(2)
with the following real and imaginary part:
(3)
(4)
For low and high frequencies, the impedance becomes real.
The above-mentioned (2) would represent a semicircle in a
complex frequency locus plot. Since real measurements show
a depressed semi-circle, a heuristic factor as a tissue-specic
dispersion parameter was introduced. In that way, the electrical
behavior of biological tissue can be described completely by the
Cole equation [7]
(5)

Here, Sramek approximated the distance of the electrodes

. In addition, the factor
as 17% of the patient's height
is the actual weight divided by the ideal weight and was introduced by Bernstein [9].
is the left ventricular ejection time (B-X interval in Fig. 2),
the static base impedance
and
the maximum of the impedance's temporal
derivative. Other models provide other equations, but all have
in common that they depend on
and
interindividual characteristics [10]. Note that since only the absolute value is measured at a xed frequency, the expression for
the static impedance is
(7)

C. Electrodes
Electrodes are transducers between electron-based charge
transports in conductors and ionic charge transports in the body.
Attaching electrodes to the body, a half cell is formed with the

414

IEEE TRANSACTIONS ON BIOMEDICAL CIRCUITS AND SYSTEMS, VOL. 9, NO. 3, JUNE 2015

Fig. 3. Equivalent circuit of a textile electrode-electrolyte skin interface,

adapted from [12].

electrode material as one part of the half cell, and electrode

gel and tissue representing the electrolyte [11]. Within this half
cell, a so-called electrical double layer generates an electrode
potential
. In the equivalent circuit of the electrode (Fig. 3),
represents the double layer,
the charge transfer
resistance which occurs at the interface, and
the resistance
of the electrolyte and the leads.
The other parts result from structures inside the tissue. The
impedance for the whole circuit is given by

(8)
In comparison to standard metal electrodes, textile electrodes
at the interface due to
show a strong capacitive behavior
the lack of hydrogel which is often used as electrolyte layer.
Sweat and skin humidity result in an additional conductive
path
in parallel to the capacitance. Thus, for textile
electrodes an equivalent circuit can be assumed (Fig. 3) and the
total impedance for the equivalent circuit can be calculated as
follows [12], [13]

(9)
In practice, two different measurement methods are used
when measuring bioimpedance by attaching electrodes on the
skin: bipolar and tetrapolar measurements. For bipolar measurements, the same electrode pair is used for current injection
as well as voltage measurement. That way,
for standard
electrodes [see (8)] or
[see (9)] will contribute twice to the measurement result since the current ow
through the electrodes causes a voltage drop over the interface
impedance. To avoid this, since only the tissue under test is
intended to be measured, tetrapolar measurements are preferred
using a second electrode pair for voltage measurements only
(Fig. 4).

Fig. 4. Tetrapolar bioimpedance measurement method.

Since operational ampliers are used to assess the voltage

drop, practically almost no current (
pA) ows through
the voltage measuring electrode leads which can, therefore, not
be externally polarized. Thus, the complex electrode interface
impedance can be excluded from the measurement.
III. SIMULATOR SETUP
To characterize bioimpedance measurement instruments
using the tetrapolar measurement method, a simulator prototype was built. The main requirements of the prototype are as
follows:
First, the complex thoracic impedance as described by the
Cole model should be implemented [see (5)]. Here, factor is
neglected since it is not necessarily needed to estimate
and
which are obligatory parameters for BIS devices to calculate, e.g., the body water content. To represent a standard human
male, the following values were implemented:
and
nF.
Second, the real part and the imaginary part of the impedance
should change over time. For most ICG devices, it would have
been sufcient to implement a dynamic resistive behavior because only the absolute value of the impedance is assessed.
However, since at least one device explicitly analyzes the imaginary part of the impedance to estimate hemodynamic parameters (the so-called bioreactance method), the simulator must be
capable of creating this signal as well [14].
Fig. 5 shows the electrical equivalent circuit for a tetrapolar
measurement. Here, the complex dynamic Cole circuit was implemented by using a variable resistor in parallel to a xed resistor in series with a capacitor.
Since
contributes to the real and the imaginary part
of the complex Cole circuit [see (4)], only
has to be altered to implement a dynamic real and imaginary part of the
impedance. In practice, this was realized using 12 parallel
single-pole double-throw (SPDT) analog solid state switches
resulting in
possible resistor combinations with a
resolution of 0.3 m . For the high precision resistors (
%
to 1%, depending on the resistance value), 12 standard values
were chosen: 1000, 200, 100, 50, 20, 10, 5, 2, 1, 0.5, 0.3 and
0.18 . The appropriate resistor combination was calculated
with Matlab on the basis of a measured impedance curve of a

ULBRICH et al.: A THORAX SIMULATOR FOR COMPLEX DYNAMIC BIOIMPEDANCE MEASUREMENTS

Fig. 7. Schematic of

415

Fig. 5. Electrical equivalent circuit of the simulator.

Fig. 8. ECG circuit.

Fig. 6. Schematic of the complex dynamic Cole impedance.

standard male human. As switches, TS3A24159 from Texas

Instruments were used. These switches have a turn-on time of
20 ns and a turn-off time of 12 ns making them fast enough
to produce high resolution ICG curves. Their digital interface
operates from 1.65 V to 3.6 V which makes them controllable
by microcontrollers. Using 3.3 V, the on-state resistance is
between 0.26
and 0.3 . The maximum continuous current of 100 mA is 10 times higher than current sources of
bioimpedance devices are allowed to apply. Fig. 10 shows the
schematic of the complex dynamic Cole impedance composed
of the Cole circuit and double-throw switches with implemented resistances for
.
Third, the base impedance
should be adjustable to simulate humans of differing height or amount of fat tissue. Therefore, in series with the Cole circuit, 4 parallel switches were
implemented for increasing
[see (7)] by
, 8.2,
18 and 27 . Hence, the new static impedance of the thorax
is
. Thus, base impedances between 24.6 and 51.6 can be simulated covering the normal
impedance range for adults. In addition, this covers the range
of clinical importance since a lowered thorax impedance due
to pulmonary edema is largely included [15]. Fig. 7 shows the

schematic of
composed of two double-throw switches and
implemented resistances.
in combination with the Cole circuit represent the frequency dependent complex dynamic impedance of the thorax.
Note that changes in
are much larger than changes in the
impedance during the heartbeat
.
Fourth, in parallel to this body impedance, an ECG has to be
created since ICG devices need this signal to calculate certain
parameters such as the pre-ejection period (PEP). This was implemented using a look-up table created from a measured ECG
in combination with the microcontroller's 12 bit digital/analog
output. The ECG generation circuit is shown in Fig. 8 without
the impedance circuit.
Since the microcontroller produces voltages up to 3.3 V
, and maximum ECG values are between 1 mV and
2 mV, a voltage divider was implemented by
and
. To
avoid a current ow through the voltage divider caused by the
bioimpedance device, since this resistance path is not considered to be part of the overall bioimpedance circuit, operational
ampliers were inserted between voltage divider and thoracic
impedance. In addition, the ampliers are serving as a buffer
to provide a precise ECG voltage level. The temporal shift
between ECG and ICG signal was physiologically matched
according to the cardiac phase.
Fifth, up to 8 complex electrode interface impedances [see
(9)] should be implemented since most ICG devices rely on 2
tetrapolar measurement leads.
and
were neglected and the interface impedances were calculated using one
bipolar
and one tetrapolar
measurement. The bipolar
measurement is necessary to measure the interface impedances
of both electrodes and the tetrapolar measurement assesses only
the impedance of the object of interest. Hence, subtracting both
measured impedances and dividing the resulting impedance by
2 calculates the interface impedance of one electrode

416

IEEE TRANSACTIONS ON BIOMEDICAL CIRCUITS AND SYSTEMS, VOL. 9, NO. 3, JUNE 2015

TABLE I
TEXTILE ELECTRODE CHARACTERISTICS

TABLE II
ESTIMATED MODEL PARAMETERS FOR TEXTILE ELECTRODE-SKIN
CONTACT IMPEDANCES

Fig. 10. Schematic of one part of the tetrapolar electrode-skin contact

impedances.

Fig. 9. Measured and modeled textile electrode-skin contact impedance.

. The measurements were made using a commercially available BIS device on a male human with three
kinds of textile electrodes from Clothing Plus, Finland.
As can be seen in Table I, all electrodes had the same size;
however, because they were made of different materials the material thickness varies resulting in different material weights.
Fig. 9 shows the measured and modeled contact impedance of
one exemplary textile.
Interestingly, these contact impedances does not reect model
assumptions found in literature. On the one hand, the contact impedances do not tend to zero for high frequencies which would
be the case for a textile contact impedance as shown in Fig. 3.
On the other hand, a standard RC element would represent a
semi-circle in the negative half-plane of the frequency locus
plot. Here, we have a deviant semi-circle which would correspond to the factor known from biological materials or tissues.
These ndings lead to a Cole-like impedance model for textile
electrode-skin contact impedances and (9) must be modied

(10)
Hence, this impedance is represented by a resistor
nected in parallel with a resistor
and a capacitor
estimated values for this model are listed in Table II.

con. All

Four possible electrode congurations were implemented: no

interface impedance (ideal electrode), and textile electrodes. In
Fig. 10, a schematic of one part of the tetrapolar contact impedances is depicted, corresponding to one current injecting electrode and one voltage measuring electrode (
in Fig. 5). It
is composed of three double-throw switches and implemented
complex contact impedances.
The circuit is powered with 12 VDC which is converted
to the required voltages. A dual layer PCB was used with
the microcontroller MSP430F1612 from Texas Instruments
soldered on the bottom layer. The microcontroller is supplied
by a 8 MHz clock and used to generate the ECG signal and to
control the switches for the dynamic impedance signal generation. All switches necessary to form an impedance are switched
at the same time within 155 ns. Fig. 11 shows a diagram of the
implemented printed circuit board (PCB) with its functional
blocks including communication paths and connections.
For the communication with a PC, a USB-connection was
realized using an FTDI FT232R. A JTAG interface was implemented to ash the microcontroller. 6 I/O ports serve as
switching lines for the electrode-skin interface impedances
at each side of the thorax (BUS 1 and BUS 2). One pair of
current injecting and voltage measuring electrodes is connected
via the
line with the thorax impedance while the
other pair uses the
line. The complex dynamic Cole
impedance is controlled by 12 I/O ports (BUS 4) while for
, 4 switching lines are necessary (BUS 4). The complex
Cole impedance is connected via the
line with
,
forming the complex dynamic thoracic impedance
.

ULBRICH et al.: A THORAX SIMULATOR FOR COMPLEX DYNAMIC BIOIMPEDANCE MEASUREMENTS

417

Fig. 12. Complex impedance measured with the Xitron, SFB7 and BCM.

TABLE III
MEASUREMENT RESULTS: BIS

Fig. 11. Diagram of the simulator hardware.

IV. MEASUREMENT RESULTS

The static complex impedance of the simulator for frequencies between 5 kHz and 1 MHz was assessed using three different measurement instruments from different manufacturers:
XiTRON Hydra 4200, XiTRON Technologies, San Diego,
USA
SFB7, ImpediMed, Pinkenba, Australia
Body Composition Monitor (BCM), Fresenius Medical
Care, Bad Homburg, Germany
The measured impedance for all devices is shown in the frequency locus plot in Fig. 12.
Interestingly, the BCM and SFB7 measure the same impedances, but the Xitron device shows a strong inductive behavior
for higher frequencies and thus deviates from the other curves.
All measured curves could be used to extract standard parameters by applying a Nelder-Mead t of the Cole model to the
measured data. The sum of squared errors of prediction (SSE)
of the t is also given in Table III.
As intended, all measurement instruments measure nearly the
same values and show that the complex impedance curve is a
semi-circle
. Apart from that, the values are in the range
of normal thoracic impedances. The mean relative error of the
extracted parameters is small (between 0.5% and 6%) because
the quality of the curve t determines how exact the Cole parameters are estimated and, in this case, the achieved SSEs do
not exceed 1.39, which is below errors estimated in real measurements
[16]. Since BIS devices do
not provide high resolution impedance measurements, they can

not assess the very small impedance changes occurring during

a heart beat.
ECG, ICG and the dynamic impedance were assessed using
3 different measurement instruments from different manufacturers:
Non-invasive continuous cardiac output monitor
(Niccomo), medis. GmbH, Ilmenau, Germany
Philips ICG, Philips Medical Systems, Andover, MA, USA
Sensatron, Philips Research, Eindhoven, The Netherlands
Measurement leads are connected to standard ECG snaps
which t all measurement instruments and which are integrated
in a proper housing. Since it is not possible to transfer raw data
from the Philips ICG device, the results were only analyzed
visually, showing that all parameters could be extracted from
the ICG curve.
Niccomo and the Philips ICG are commercially available devices, whereas the Sensatron is a new development from Philips
[17]. It is a small wearable datalogger to assess not only ICG
and ECG, but also PPG and heart sounds. Since the device is
still under development, the measured data are not yet calibrated
(Fig. 13).
The impedance was ltered using a digital zero-phase 5th
order Butterworth bandpass lter (0.8 Hz
Hz) in
Matlab. In impedance cardiography, it is common to lter and

418

IEEE TRANSACTIONS ON BIOMEDICAL CIRCUITS AND SYSTEMS, VOL. 9, NO. 3, JUNE 2015

Fig. 15. Static transfer behavior (simulated versus measured impedance).

Fig. 13. ECG, Impedance and ICG measured with the Sensatron.

Fig. 14. Calculated impedance by Matlab, unltered impedance (dashed)

measured with the Niccomo and ltered signal (dotted).

Fig. 16. ECG, Impedance and ICG measured with the Niccomo.

average the signal to cancel thermal noise or respiratory and motion artifacts, since this is necessary to determine the characteristic points of the ICG curve. In commercially available devices,
this is done automatically before calculating the SV [18], [19].
Fig. 14 visualizes the very good match of the measured
unltered impedance and the original preset impedance since
the peak-to-peak impedances match perfectly, the RMS error
is 0.0064 , and the peak value of the cross-correlation is
, which shows that the morphology is nearly identical. The remaining ripple is an inherent error of the switches,
which have an on-state resistance between 0.26 and 0.3 ,
producing an error of 0.04 in the worst case. Here, this ripple
is negligible due to the subsequent ltering of the ICG curve
before characteristic point extraction. Thus, none of the characteristic points are shifted. The only difference is a decrease
of the ICG maximum (0.02 /s) leading to a SV difference of

1 ml. This is a difference far below the SV difference produced

by the choice of characteristic point extraction algorithms [20].
Typical values would be a LVET difference of 15.5 ms leading
to a SV difference of 5.2 ml in our case. Note that the offset
was subtracted from the impedance curves to facilitate their
comparison. An evaluation of the offset error is given by the
spectroscopical analysis (see Table III).
The linear curve progression of the static transfer behavior
(Fig. 15) shows that the simulator correctly produces the desired
impedance changes.
Comparison of ECG, Impedance and ICG measured with the
Niccomo (Fig. 16) and the Sensatron (Fig. 13) shows that both
devices measure the same curves. The peak-to-peak impedance
is in the range of normal physiological values.
The ICG curve was used to calculate common hemodynamic
parameters by the measurement device to show that the results

ULBRICH et al.: A THORAX SIMULATOR FOR COMPLEX DYNAMIC BIOIMPEDANCE MEASUREMENTS

TABLE IV
MEASUREMENT RESULTS: ICG

reect realistic values measured on a human thorax. These parameters are heart rate (HR), stroke volume (SV), pre-ejection
period (PEP), left ventricular ejection time (LVET) and cardiac
output (CO). For this work, we used the Bernstein-Sramek equation to calculate the SV since the Niccomo uses the same equation internally. The results are shown in Table IV.
The implemented impedances, as described in Section III, do
not affect the measurements, as stated in Section II.C, if the current source of the measurement device was designed correctly.
Hence, the same measurement results were obtained with all devices, independent from the type of electrode.
V. DISCUSSION
Only three works are known to the authors that deal with
the development of ICG simulators for testing and calibration
of ICG devices. All works have in common, that they show
no physiologic impedance curves produced by the simulator.
Only one paper describes a rectangular impedance change used
for the simulation of an electroglottographic signal [3]. However, a rectangular signal would be not sufcient for testing of
ICG devices since a rectangular impedance change would produce dirac-like pulses for rising and falling edges. This temporal derivative could not be analyzed to extract characteristic
points. Unfortunately, no minimal quantization is mentioned in
this paper so that no statement can be given if the circuit would
basically be capable of producing realistic impedance changes
for ICG. But since this circuit does not contain an ECG generator, it does not seem to be developed to serve as an ICG
simulator.
A second work claiming to present an ICG simulator rather
introduced a square impedance generator with an amplitude
of 0.2
[4]. Another work which describes a simulator
containing an ECG generator and aiming to develop an ICG
simulator seems to follow the same approach: the minimal
resolution of the simulated resistance is so small that only
steps of 0.0237
can be realized using the lowest possible
base impedance. The dynamic impedance was implemented as
[2]. These steps would also cause
an ICG device to produce only uninterpretable derivatives of
the measured impedance.
Another commonality of all works is that they implemented
only resistors so that all capacitive effects occuring inside and
outside the body are neglected. One the one hand, this restricts
the simulator to be used for ICG devices only. Neither bioreactance devices nor BIS devices can be tested. On the other hand,
pure resistive changes can not reect the capacitive load of the

419

human body (Cole model) or even more importantly the body

in combination with textile electrodes. This is a big drawback
since the capacitive load is the major reason for bioimpedance
devices to fail when using textile electrodes especially when a
broad frequency range must be supplied as it is done in BIS
devices.
Concerning the implemented electrode-skin contact impedances, the existing model assumptions in literature should be
extended by an additional resistance and a factor which takes
a deviant semi-circle into account. This model would be more
representative for measured impedances. The explanation of the
need for such an extension is that old models assume air inclusions between textile material and skin due to the surface texture
of the textile electrode. In reality, these inclusions do not exist
since textile materials are designed in a manner that they are
exible enough to adapt to the skin surface. In addition, these
tiny gaps, if they exist, are quickly lled with sweat acting as a
conductive electrolyte.
VI. CONCLUSION
In this work, a thorax simulator was developed to test and
calibrate bioimpedance devices. It provides not only the complex frequency dependent thoracic impedance for BIS devices,
but also a dynamic complex impedance and an ECG for ICG
devices. In addition, the thoracic base impedance can be altered
as well as the complex electrode-skin interface impedances for
up to 8 electrodes.
It was shown that the device simulates physiologic impedances which match very well with the calculated preset values
reecting realistic measurements on the human thorax.
In addition, new developments, such as the Philips Sensatron
or the Quadra device from Tallinn University of Technology
(Estonia) which measures the complex impedance continuously
between 10 mHz to 500 kHz, can be tested with this simulator [21].
For future developments, pathologies such as lung edema
should be taken into account since they can have a considerable
inuence on the signal. Recent nite element method simulations have shown that a signal amplitude decrease of over 50%
is possible due to lung edema in heart failure patients [22].
In practice, electrode interface impedances are not always
stable. In particular when focusing on wearable devices in personal health care scenarios, these impedances change due to motion and produce artifacts in the measured signal. This occurs
for standard adhesive electrodes and gets worse when clothing
with textile electrodes is worn, since the distance between skin
and electrode can vary even more. To analyze the ability of measurement instruments to cope with these sources of error, artifact
generation will be implemented as well.
REFERENCES
[1] Bundesministerium fr Bildung und Forschung Gesundheitsforschung hautnah erleben, 2007 [Online]. Available: http://www.
bmbf.de/press/2171.php
[2] B. B. Patil, V. K. Pandey, and P. C. Pandey, A microcontroller
based thorax simulator for testing and calibration of impedance
cardiographs, in Proc. Int. Symp. Emerging Areas in Biotechnology
& Bioengineering, Mumbai, India, Feb. 2009.
[3] V. K. Pandey, P. C. Pandey, and J. N. Sarvaiya, Impedance simulator
for testing of instruments for bioimpedance sensing, IETE J. Res., vol.
54, no. 3, pp. 203207, 2008.

420

IEEE TRANSACTIONS ON BIOMEDICAL CIRCUITS AND SYSTEMS, VOL. 9, NO. 3, JUNE 2015

[4] N. S. Manigandan, V. K. Pandey, and P. C. Pandey, Thoracic simulator for impedance cardiography, in Proc. Nat. Symp. Instrumentation, Pantnagar, India, Nov. 2003.
[5] International Electrotechnical Comission, IEC 60601-1-1: Medical
electrical equipment, 2000.
[6] H. Schwan, Electrical properties of tissue and cell suspensions, Adv.
Biol. Med. Phys., vol. 5, pp. 147209, 1957.
[7] S. Grimnes and O. G. Martinsen, Bioimpedance and Bioelectricity Basics, 2nd ed. Amsterdam, The Netherlands: Elsevier, 2008.
[8] T. Debski, Y. Zhang, J. R. Jennings, and T. W. Kamarck, Stability of
cardiac impedance measures: Aortic opening (b-point) detection and
scoring, Biol. Psych., vol. 36, pp. 6374, 1993.
[9] J. M. V. D. Water, T. W. Miller, R. L. Vogel, B. E. Mount, and M. L.
Dalton, Impedance cardiographyThe next vital sign technology?,
Chest, vol. 123, pp. 20282033, 2003.
[10] D. P. Bernstein and H. J. M. Lemmens, Stroke volume equation
for impedance cardiography, Med. Biol. Eng. Comput., vol. 43, pp.
443450, 2005.
[11] L. A. Geddes and L. E. Baker, Principles of Applied Biomedical Instrumentation. New York, NY, USA: Wiley, 1968.
[12] L. Beckmann, C. Neuhaus, G. Medrano, N. Jungbecker, M. Walter,
T. Gries, and S. Leonhardt, Characterization of textile electrodes and
conductors using standardized measurement setups, Phys. Meas., vol.
31, pp. 233274, 2010.
[13] T. InOh, S. Yoon, T. E. Kim, H. Wi, K. J. Kim, E. J. Woo, and R.
Sadleir, Nanober web textile dry electrodes for long-term biopotential recording, IEEE Trans. Biomed. Circuits Syst., vol. 7, no. 2, pp.
204211, Apr. 2013.
[14] P. Squara, Bioreactance: A new method for non-invasive cardiac
output monitoring, Intensive Care Med., pp. 619630, 2008.
[15] L. A. H. Critchley, Impedance cardiography: The impact of new technology, Anaesthesia, vol. 53, pp. 677684, 1998.
[16] L. Ward, B. H. Cornish, N. I. Paton, and B. J. Thomas, Multiple
frequency bioelectrical impedance analysis: A cross-validation study
of the inductor circuit and cole models, Phys. Meas., vol. 20, pp.
333347, 1999.
[17] J. Muehlsteff, P. Carvalho, J. Henriques, R. P. Paiva, and H. Reiter,
Cardiac status assessment with a multi-signal device for improved
home-based congestive heart failure management, in Proc. Annu. Int.
Conf. IEEE Engineering in Medicine and Biology Soc., Aug. 30-Sep. 3,
2011, pp. 876879.
[18] W. Hu, H. H. Sun, and X. Wang, A study on methods for impedance
cardiography, in Proc. 19th Annu. Int. IEEE Conf. Engineering in
Medicine and Biology Soc., 1997, vol. 5, pp. 20742077.
[19] O. Dromer, O. Alata, and O. Bernard, Impedance cardiography ltering using scale fourier linear combiner based on rls algorithm, in
Proc. Annu. Int. Conf IEEE Engineering in Medicine and Biology Soc.,
Sep. 2009, pp. 69306933.
[20] P. Carvalho, Robust characteristic points for ICG: Denition and comparative analysis, in Proc. Int. Conf. Bio-Inspired Systems and Signal
Processing, 2011.
[21] M. Min, QuadraImpedance Spectroscopy Device for Dynamic Measurements, 2012.
[22] M. Ulbrich, J. Mhlsteff, M. Walter, and S. Leonhardt, Simulation of
lung edema in impedance cardiography, Comput. Cardiol., vol. 39,
pp. 3336, 2012.

Mark Ulbrich was born in Bonn, Germany, in

1980. He studied electrical engineering with a focus
on medical engineering and received the Dipl.-Ing.
degree from RWTH Aachen University, Aachen,
Germany, in 2009.
Currently, he is working toward the Ph.D. degree
at the Chair for Medical Information Technology
(MedIT), RWTH Aachen University, where he is
a Research Assistant. His research interests are in
the eld of impedance cardiography, particularly
nite element simulations of electromagnetic elds,
hardware development, textile integration of electronics, and clinical trials.

Jens Mhlsteff was born in Neuhaus am Rennweg,

Germany, in 1971. He received the M.Sc. degree in
physics from the University of Jena, Jena, Germany,
and the Ph.D. degree from the University of the
German Federal Armed Forces, Munich, Germany,
in 1998 and 2002, respectively.
He has worked on biomedical sensors for monitoring solutions in personal health care applications
since joining Philips Research in 2002. He is a
Senior Scientist in the Patient Care Solutions Group,
where he leads projects on novel biomedical sensor
research.

Daniel Teichmann (S'12) was born in Essen, Germany, in 1982. He received the Dipl.-Ing. degree in
electrical engineering from RWTH Aachen University, Aachen, Germany.
Currently, he is working toward the Dr.-Ing.
(Ph.D.) degree at the Chair of Medical Information
Technology, RWTH Aachen University, where he is
a Research Assistant. His research interests include
physiological measurement techniques and signal
processing.

Steffen Leonhardt (M'95SM'06) was born in

Frankfurt, Germany, in 1961. He received the
M.S. degree in computer engineering from SUNY,
Buffalo, NY, USA, the Dipl.-Ing. degree in electrical
engineering and the Dr.-Ing. degree in control
engineering from the Technical University of Darmstadt, Darmstadt, Germany, and the M.D. degree in
medicine from J. W. Goethe University, Frankfurt,
Germany.
He has ve years of R&D management experience
in medical engineering industry and was appointed
Full Professor and Head of the Philips endowed Chair of Medical Information
Technology at RWTH Aachen University, Aachen, Germany, in 2003. His research interests include physiological measurement techniques, personal health
care systems, and feedback control systems in medicine.

Marian Walter (SM'13) was born in Saarbrcken,

Germany, in 1966. He studied electrial engineering,
with a specialization in control engineering, at Technische Universitt Darmstadt, Darmstadt, Germany,
from which he received the Dipl.-Ing. and Dr.-Ing.
degrees in 1995 and 2002, respectively.
He has worked in the medical engineering industry for three years and was appointed Senior
Scientist at the Philips Chair of Medical Information
Technology at RWTH Aachen University, Aachen,
Germany, in 2004. His research interests include
non-contact monitoring techniques, signal processing, and feedback control in
medicine.