Thoracic Imaging
David F. Yankelevitz, MD
Anthony P. Reeves, PhD
William J. Kostis, MS
Binsheng Zhao, DSc
Claudia I. Henschke, PhD,
MD
Index terms:
Computed tomography (CT),
comparative studies, 60.12115,
60.12117, 60.12118
Computed tomography (CT),
experimental studies, 60.12115,
60.12117, 60.12118
Computed tomography (CT), thin-
section, 60.12118
Lung, nodule, 60.281
Lung neoplasms, CT, 60.12115,
60.12117, 60.12118
Lung neoplasms, diagnosis, 60.281,
60.30
Radiology 2000; 217:251256
Abbreviations:
2D two-dimensional
3D three-dimensional
1
From the Department of Radiology,
New York Presbyterian Hospital-Weill
Cornell Medical Center, Building Starr-
8A23, 525 E 68th St, New York, NY
10021 (D.F.Y., B.Z., C.I.H.) and the
School of Electrical Engineering, Cor-
nell University, Ithaca, NY (A.P.R., W.J.K.).
From the 1998 RSNA scientific assembly.
Received May 13, 1999; revision re-
quested July 22; final revision received
January 20, 2000; accepted February
1. D.F.Y., A.P.R., W.J.K., and C.I.H.
supported in part by National Insti-
tutes of Health grants R01-CA-63393
and R01-CA-525928. Address corre-
spondence to D.F.Y. (e-mail: dyankele
@mail.med.cornell.edu).
RSNA, 2000
Author contributions:
Guarantors of integrity of entire study,
D.F.Y., A.P.R., C.I.H.; study concepts
and design, D.F.Y., A.P.R., C.I.H.; def-
inition of intellectual content, D.F.Y.,
A.P.R., C.I.H.; literature research, D.F.Y.,
C.I.H.; clinical studies, D.F.Y., C.I.H.; ex-
perimental studies, D.F.Y., A.P.R., W.J.K.;
data acquisition, all authors; data analy-
sis, D.F.Y., A.P.R., W.J.K., C.I.H.; manu-
script preparation, editing, and review,
all authors
Small Pulmonary Nodules:
Volumetrically Determined
Growth Rates Based on CT
Evaluation1
PURPOSE: To determine the accuracy of high-resolution computed tomographic
(CT) volumetric measurements of small pulmonary nodules to assess growth and
malignancy status.
MATERIALS AND METHODS: The accuracy of three-dimensional (3D) image
extraction and isotropic resampling techniques was assessed by performing three
experiments. The first experiment measured volumes in spherical synthetic nodules
of two diameters (3.20 and 3.96 mm), the second measured deformable silicone
synthetic nodules prior to and after their shape had been altered markedly, and the
third measured nodules of various shapes and sizes. Three-dimensional techniques
were used to assess growth in 13 patients for whom the final diagnosis was known
and whose initial nodule diameters were less than 10 mm. By using the exponential
growth model and the calculated nodule volume at two points in time, the doubling
time for each subject was calculated.
RESULTS: The three synthetic nodule studies revealed that the volume could be
measured accurately to within 3%. All five malignant nodules grew, and all had
doubling times less than 177 days. Some malignant nodules had asymmetric
patterns of growth identified by using the 3D techniques but not the two-dimen-
sional methods. All eight benign nodules had doubling times of 396 days or greater
or showed a decrease in volume.
CONCLUSION: CT volumetric measurements are highly accurate for determining
volume and are useful in assessing growth of small nodules and calculating their
doubling times.
Determination of growth rates for a pulmonary nodule can be of vital importance in light
of the high mortality associated with lung cancer. With ever-improving resolution and
availability of computed tomographic (CT) scanners, an increasing number of smaller
nodules are being detected (1,2). The challenge for radiologists is to help in the noninva-
sive differentiation between benign and malignant nodulesin other words, to estimate
their probability of malignancy. Computer-aided methods are now being developed to aid
the radiologist in both the detection and the diagnosis of pulmonary nodules (37).
When small nodules are detected incidentally on CT scans, no standard approach for
diagnostic work-up currently exists, to our knowledge. All of the existing diagnostic
techniques for determining their cause become less accurate with decreasing nodule size.
Although it is recognized that the majority of small solitary nodules are benign (8 10),
suggestions for further work-up range from follow-up with chest radiographs or CT to a
variety of additional diagnostic studies, to semiinvasive procedures, to surgical resection.
Temporal information gained by comparison with prior studies can provide important
information as to the stability or growth of the nodule. Thus, accurate measurements of
change on repeat imaging studies provide a reasonable, noninvasive technique for pre-
diction of malignancy (11). In fact, stability in nodule size during a 2-year period generally
is accepted as sufficient in lowering the probability of malignancy to make the perfor-
mance of additional diagnostic tests unnecessary (12,13).
251
 We previously showed that high-reso-
lution (thin sections and small field of
view) CT imaging could be used to assess
nodule growth (6). In this previous study,
we selected the CT image containing the
largest cross-sectional area of the nodule
and determined the nodule area for each
point in time. Using the change in the
nodule area, we calculated the doubling
time by using an exponential growth
model. This provides a conservative esti-
mate of growth. We showed that malig-
nant nodules all demonstrated growth
rates consistent with malignant tumors,
whereas benign nodules did not (6).
The purpose of this study was to deter-
mine the accuracy of high-resolution CT
volumetric measurements of small pul-
monary nodules to assess growth and
malignancy status over time. This ap-
proach has substantial advantages, be-
cause the determination of volume is less
dependent on patient positioning and
section selection and allows for detection
of asymmetric growth.
MATERIALS AND METHODS
The use of three-dimensional (3D) tech-
niques for nodule volume determination
requires contiguous high-resolution CT
images of the entire nodule. Further-
more, since our focus was on small nod-
ules, it was important to achieve the
highest possible resolution along the
scan axis by using the smallest pitch pos-
sible. Thus, our image data were acquired
by using a CT scanner (HiSpeed Advan-
tage; GE Medical Systems, Milwaukee,
Wis) in helical mode at 1:1 pitch, 140
kVp, and 200 mA by using 1-mm beam
collimation. The whole nodule region
was scanned in a single breath hold. The
images were reconstructed at 0.5-mm in-
tervals with a 9.6-cm field of view by
using the high-spatial-resolution algorithm.
Using a software package (VISIONX; Cor-
nell University, Ithaca, NY; available at:
http://www.ee.cornell.edu/reeves/research
/visx/index.html) for multidimensional im-
age processing and computer vision, the
radiologist (D.F.Y.) selected a region of
interest containing the nodule. The im-
age sections containing the region of in-
terest were then extracted to obtain the
reconstructed 3D image volume of the
region of interest. This was accomplished
by resampling the 3D image to isotropic
(resolution is identical in all three dimen-
sions) space and then thresholding to ob-
tain a 3D binary image, followed by the
application of the 3D segmentation tech-
niques used to define the nodule region.
252 Radiology October 2000
All size measurements were made on this
filtered, segmented nodule image. This
technique was used for segmentation of
both the synthetic and in vivo nodules.
We have developed an interactive
graphical user interface, or GUI, that per-
mits the radiologist to set parameters and
to depict the results at each step of this
process. The system also allows for visual
comparisons of the nodule at all stages of
the analysis.
To help resolve partial volume effects
and obtain subvoxel accuracy, the data
were resampled to an isotropic space by
Figure 1. Three-dimensional shaded-surface
using three-dimensional linear (trilinear) images from high-resolution CT scans of two
interpolation (14). This produced an im- deformable silicone synthetic nodules. Each
age having higher resolution than the pair of images shows a single nodule. The im-
original image on the basis of a 0.25-mm3 age on the left was scanned first; then, after
supergrid. In the new, isotropic space, we deforming the nodule, the image on the right
was obtained. These synthetic nodules span
applied a series of 3D morphologic filters
the size range that was tested. The top nodule
to help in segmenting the nodule from has an initial maximum diameter of approxi-
other adjacent objects, such as vessels mately 7 mm, and the bottom nodule has an
and bronchi, or from the pleural surface, initial maximum diameter of approximately
while preserving the relevant nodule 14 mm.
characteristics.
Once the nodule was segmented, vol-
ume was determined by adding the num-
in the volume or area was made, we as-
ber of voxels contained in the resultant
sumed that the nodule growth or de-
image. Similarly, the two-dimensional
crease could be represented by the simple
(2D) area was determined by counting
exponential growth model. This model
the number of pixels contained in the
allows us to calculate the tumor doubling
single image having the largest cross-sec-
time. Doubling time (DT) may be ex-
tional area (6). The same segmentation
pressed in terms of direct volume (V)
techniques were applied to all repeat im-
measurements as DTV [log 2 t]/[log
ages of the nodule acquired at different
(V 2 /V 1 )], where t is the change over
times to minimize additional segmenta-
time. With the use of 2D diameter esti-
tion errors; for purposes of growth assess-
mates (D), doubling time may be calcu-
ment, it is more important to precisely
lated as DTD [log 2 t]/[3log (D 2 /
determine the relative change in nodule
D 1 )]. By using 2D area estimates ( A),
volume than the absolute volume mea-
doubling time may be calculated as
surements. A full description of these tech-
DTA [2log 2 t]/[3log ( A 2 /A 1 )]. It
niques is beyond the scope of this article.
is important to note that the doubling
To compare the accuracy of the 2D and
time calculation based on 2D diameter
3D techniques, we determined the nod-
estimates assumes uniform growth in three
ule diameter, D, on the basis of the nod-
dimensions of a spherical object. Since
ule area and volume. In calculating D on
nodules do not necessarily grow uni-
the basis of the 2D nodule area, we as-
formly in all dimensions, true volumetric
sumed that the nodule was a perfect cir-
determinations theoretically should pro-
cle. Similarly, to determine D on the ba-
vide more accurate information.
sis of the nodule volume, we assumed
that the nodule was a perfect sphere.
Synthetic Nodule Data
Since we were most interested in the
consistency of our technique and the A series of three experiments was per-
ability to detect relative change, we cal- formed by using synthetic nodules. In
culated the percentage of error of each each of these, the nodules were placed on
measurement relative to the area or vol- synthetic foam having an attenuation
ume. In this context, we defined the per- similar to that of aerated lung. Related
centage of error as the mean-normalized work in measurement of synthetic nod-
square root of the mean-squared error rel- ules has been described (15). These exper-
ative to each metric. This measurement iments were designed to test the preci-
also can be thought of as the coefficient sion and reproducibility of our scan and
of variation, or SD normalized by the image processing techniques in making
mean, of each calculated metric. volumetric determinations. The clini-
Once the determination of the change cally relevant pulmonary nodules were
Yankelevitz et al

TABLE 1
Diameter Estimates and Resultant Percentages of Error Calculated from
the Area and Volume of Calibrated Synthetic Spherical Nodules
on High-Resolution CT Images
Basis for Diameter Estimation*
Synthetic Spherical Nodules 3D Volume 2D Area
Larger diameter (3.96 mm)
1.0 3.96 0.012 (0.70) 3.78 0.010 (0.63)
0.5 3.96 0.012 (0.69) 3.78 0.011 (0.67)
0.2 3.96 0.011 (0.67) 3.79 0.009 (0.58)
Smaller diameter (3.20 mm)
1.0 3.14 0.017 (1.47) 2.95 0.018 (1.67)
0.5 3.15 0.017 (1.46) 2.96 0.014 (1.32)
0.2 3.15 0.016 (1.43) 2.98 0.013 (1.23)
* Data are the diameters in millimeters plus or minus SD. Data in parentheses are percentages
(root-mean-square volume error).
Data are the reconstruction intervals in millimeters.
obtaining high-resolution CT scans in 20
TABLE 2 silicone (mean attenuation, approxi-
Percentages of Error of the Volume
Measurements Obtained in Silicone mately 175 HU) synthetic nodules that
were 315 mm in diameter. Then, using
Synthetic Spherical Nodules on
the Basis of Diameter and the principle that a noncompressible liq-
Reconstruction Interval uid or solid retains the same volume even
when its shape is altered, we deformed
Root-
Mean- the nodules and rescanned them. Repre-
Reconstruction No. of Diameter Square sentative CT images obtained both before
Interval (mm) Nodules (mm) Error (%) and after deformation are shown in Fig-
ure 1. Volumes were calculated for each
0.5 19 36 1.05
16 611 0.48 nodule before and after deformation so
1.0 11 36 2.88 that the percentage of error could be de-
8 611 2.34 termined.
For the third experiment, we con-
structed 35 silicone synthetic nodules
with diameters of 311 mm. Volumetric
those smaller than 1.0 cm in diameter. scans were obtained for each of these
Nodules larger than this are evaluated nodules, and the volumes were calcu-
fairly easily by using other techniques, lated. These nodules were then weighed
including biopsy or positron emission to- on an analytical balance to a precision of
mography. Thus, our main interest was 0.1 mg. Since the density of the material
in small nodules. In general, a round or was uniform, the volume and mass should
spherical shape was sufficient in repre- be proportional for each nodule, and we
senting these small nodules. The attenu- plotted the volume versus the mass and
ation of our simulated nodules was higher calculated the percentage of error.
than that of actual clinically detected
nodules; however, attenuation differences
Patient Data
were not substantial due to the inherently
high contrast to lung parenchyma. To demonstrate the feasibility and
The first experiment involved two re- clinical utility of these volumetric tech-
peat scans of a set of 21 calibrated (1% niques, we examined 13 subjects (seven
variation) acrylic spheres, each having a men, six women; age range, 60 77 years;
diameter of 3.96 mm, and two repeat mean age, 67.5 years) whose pulmonary
scans of another set of 29 highly cali- nodules were initially less than 10 mm in
brated spheres, each having a diameter of diameter and had been imaged repeat-
3.20 mm. The mean attenuation of these edly as part of the routine work-up. The
spheres was approximately 75 HU. High- mean attenuation of these nodules was
resolution images were obtained, and im- 35 HU. Since these images were obtained
ages were reconstructed every 1.0, 0.5, or as part of routine clinical practice and
0.2 mm. The percentage of error of the not ordered specifically as part of any
area and volume for each of the two sets experimental protocol, the interval be-
of spheres was then calculated. tween the repeat scans varied (20 745
The second experiment consisted of days). In those patients in whom the de-
Volume 217 Number 1 Small Pulmonary Nodules: Volumetrically Determined Growth Rates
lay was short (less than 40 days), the sec-
ond set of images usually was obtained as
part of a CT-guided percutaneous lung
biopsy procedure. In those with longer
delays, a clinical decision was made to
follow them up by means of observation
rather than proceeding to invasive diag-
nostic procedures. It is beneficial in con-
firming the validity of our techniques
that we had some patients with long and
others with short doubling times, as well
as those with long and others with short
interscan delays. The final diagnosis in
each nodule was based on either biopsy
results or lack of growth for more than 2
years: five were malignant, and eight
were benign.
The area and volume for each nodule
was calculated for each point in time. By
using the change in the area, the diame-
ters and doubling times were calculated
for each nodule. Similarly, by using the
change in the volume, the diameters and
doubling times for each nodule were cal-
culated.
RESULTS
Synthetic Nodule Data
The results of the first synthetic nodule
experiment in which volumetric and area
measurements of spheres were compared
are shown in Table 1. Estimates of nodule
diameter and the resultant percentages of
error for the small and large spheres are
given for three reconstruction intervals.
Compared with the 1.0-mm reconstruc-
tion, the smaller ones improved the ac-
curacy of the measurements only slightly.
Also, both area and volume measurements
for the larger spheres were more consis-
tent than those for the smaller spheres.
Overall, there was little difference in the
volume measurements compared with
the area measurements. Note that all of
the volume measurements had percent-
ages of error less than 2%.
Results of the measurement of deform-
able silicone synthetic nodules at two re-
construction intervals are given in Table
2. The amount of deformation of these
nodules was considerable and was much
greater than the change that would occur
in the clinical situation, yet all repeat
volume measurements at all sizes had a
percentage of error of 0.9%2.8%.
Figure 2 illustrates the percentage of
error in volume measurements for 35
spheres reconstructed at 0.5-mm inter-
vals. In this third experiment, we found
that there was at most a 2%3% variation
between individual volume and mass
measurements, the variation being greater
253
for smaller nodules. For example, for
nodules smaller than 6 mm in diameter, TABLE 3
In Vivo Nodule Diameters and Doubling Times Based on Changes in the Volume
the root-mean-square variation was within and Area Measurements
plus or minus 1.1%; in nodules larger
than 6 mm in diameter, it was within
plus or minus 0.5% (Fig 2). The greater Volume (mm3) Area (mm2) Doubling Time
Diameter (d)
percentage of error for these smaller nod- Patient No. of Days Estimate First Second First Second Final
ules again relates to the increased propor- No. between Scans (mm) Scan Scan Scan Scan Volume Area Diagnosis*
tion of partial voxel effects on the surface 1 36 6.9 106.9 135.7 36.5 36.6 104 9,700 Malignant
of the smaller nodules. The data clearly 2 20 9.3 239.8 313.8 65.9 74.1 51 78 Malignant
illustrate the inverse relationship of mea- 3 69 5.4 141.3 184.8 18.1 25.7 177 90 Malignant
surement error to nodule size, as well as 4 71 6.5 265.2 466.4 32.6 66.9 87 46 Malignant
5 33 5.5 62.5 85.3 250.1 341.2 73 49 Malignant
our ability to detect small volume differ- 6 745 3.9 89.0 166.4 11.4 28.3 826 378 Benign
ences in nodules of similar size. 7 35 7.4 70.0 70.9 280.1 283.4 2,030 135 Benign
Results from these synthetic nodule 8 35 7.2 54.6 56.3 218.5 225.3 798 532 Benign
studies suggest that the image resolution 9 84 4.1 36.2 36.2 13.0 14.9 33,700 288 Benign
10 225 4.0 41.5 37.6 12.2 11.8 1,570 2,840 Benign
is adequate following isotropic resam- 11 61 7.1 208.6 219.3 38.9 46.3 846 164 Benign
pling to determine change in volume 12 70 8.4 207.9 222.2 52.4 53.6 731 1,520 2YNC
within a 2% error. However, when the 13 306 5.8 91.5 156.2 25.6 34.1 396 494 2YNC
shape is altered substantially (Fig 1), then * 2YNC 2 years without change.
the change in volume can be determined
within a 3% error, although this can be
reduced to 1% if the CT images are recon-
structed at 0.5-mm intervals. Further-
more, 3D volumetric measurements ap- the benign nodules had relatively short
peared to be better than or comparable to area-based doubling times: 164, 288, and
2D area measurements. 378 days for patients 11, 9, and 6, respec-
tively; the doubling times based on the
volume were 846, 33,700, and 826 days,
Patient Data respectively. Some malignant nodules
The volume and area measurements had asymmetric patterns of growth iden-
obtained from in vivo nodules are given tified by using the 3D techniques but not
in Table 3 together with the time in days the previously developed 2D methods
between the two scans. Table 3 also (eg, patient 1).
shows the estimated doubling times for
each nodule separately on the basis of DISCUSSION
volume change and area change. The fi-
nal diagnosis is shown in the last col- Three-dimensional methods for estimat- Figure 2. Scatterplot illustrates percentage of
umn: malignant, benign, or 2 years of ing growth offer intrinsic advantages error of the volume measurements in 35 spher-
follow-up without change. To appreciate over conventional 2D methods. Since the ical synthetic nodules that were 311 mm in
the approximate size of each nodule, a whole nodule is used, advantage is taken diameter and reconstructed at 0.5-mm inter-
vals. The root-mean-square (RMS) and maxi-
diameter calculation based on the initial of all available data, not just a subset
mum percentage of error for nodules 3 6 mm
volume is provided. The initial diameters (section with maximum cross-sectional in diameter are 1.1% and 2.8%, respectively.
ranged from 3.9 to 9.3 mm. area). When performing the repeat scan, For nodules 6 11 mm in diameter, these are
Doubling times based on volume mea- there is no need to select matching im- 0.5% and 0.9%. The greater variability seen in
surements were substantially smaller for ages on the follow-up study (registra- the smaller nodules reflects the increased pro-
malignant nodules than for benign nod- tion), since total volume is measured, portion of partial voxels. With increasing size,
the amount of error in volume estimation be-
ules. This difference was also seen with which makes the volumetric measure-
comes progressively smaller.
area measurements, although the differ- ments independent of registration. Reg-
ence was less clear. By using volumetric istration occasionally can cause difficulty
measurements, all malignant nodules when patients are not oriented in exactly
had doubling times of less than 177 days, the same position during repeat scans as the entire volume of the nodule must be
which is considered well within the typ- they were during the original study or, obtained in a single breath hold.
ical range for malignancy (16). All of the despite the use of thin sections, the im- In a previous study (6), we were able to
benign nodules had doubling times that ages do not correspond precisely. A final demonstrate the potential of early repeat
were 396 days or greater on the basis of advantage is the ability to view the nod- scanning with use of 2D techniques. This
volumetric measurements. When the nod- ule in 3D space from any arbitrary view- was accomplished by using both syn-
ule decreased in volume over time, as in point. This provides additional visual thetic nodules and actual patient data.
patient 10, the doubling time was nega- verification concerning the accuracy of We showed that CT has the spatial reso-
tive. the segmentation process. lution necessary for detection of change
Results of area measurement were The additional challenge in 3D volu- in volume within 30 days for nodules
more variable with one malignant nod- metric determination is that motion arti- that have doubling times within 30 180
ule (patient 1) having a long doubling facts and scanning speed introduce addi- days and are larger than 5 mm in diam-
time of 9,700 days. In addition, several of tional errors in the data acquisition. Also, eter at initial presentation, and we pre-

254 Radiology October 2000 Yankelevitz et al


Figure 3. Patient 1. High-resolution CT images of a nodule scanned at two different times; AC
show the first scan and DF show the second scan obtained 36 days after the first. A and D show
the largest area section, B and E show the selected 2D region, and C and F are 3D shaded-
surface coronal images. Although the area of the nodule seems similar on the transverse images
(A, B, D, and E), a substantial change in overall volume can be appreciated on the 3D images.
sented our results in 15 actual nodules. In
this initial study, we used k-means clus-
tering (17) as the segmentation method-
ology. As we gained further experience
with nodule growth analysis, we found it
necessary to develop other segmentation
techniques that are more responsive to
the complexities we encountered, nota-
bly separation of adjacent structures. We
have developed a variety of techniques
for this purpose that involve threshold-
ing and the application of 3D geometric
filters.
In this study, we compared the accu-
racy of 3D techniques in determining
volume relative to 2D techniques that
were used to measure cross-sectional
area. We were able to demonstrate that
volume measurements were at least as
consistent as area measurements when
evaluating consistency of measurements
of calibrated synthetic spheres. We com-
pared total volume versus maximal cross-
sectional area and showed the relative
equality of these techniques. This was
particularly important because the 3D
data were anisotropic compared with the
2D data. Thus, even in idealized circum-
stances in which synthetic spheres were
used and the superior in-plane resolution
of 2D imaging would be expected to have
its greatest advantage, 3D techniques
were at least as good. This is presumably
in part due to the inherent advantage of
making a volumetric measurement as op-
Volume 217 Number 1
maximal cross-sectional area did not
change, overall volume changed substan-
posed to a 2D measurement, which com-
pensates for the anisotropic voxel size.
Two-dimensional techniques in which
thin sections with overlapping reconstruc-
tions are used still allow for slight error
when trying to choose corresponding sec-
tions of maximal cross section. This ex-
periment also confirmed that our tech-
nique of isotropic resampling yielded
adequate image resolution.
Further evidence for the accuracy of
the 3D techniques was found in the sta-
bility of the volume measurements from
the deformable silicone synthetic nod-
ules, as well as from experiments in which
the masses of nodules of various sizes was
known precisely. All synthetic nodule
studies yielded results that were consis-
tent to approximately 3%.
In our evaluation of the in vivo nod-
ules, the doubling times estimated by
means of volumetric techniques appeared
to be more consistent with the final patho-
logic diagnosis, as opposed to those esti-
mated by means of the 2D techniques.
This series of nodules was different from
that used in the prior study (6), because
all the nodules for our current study had
to have a complete series of contiguous
images to be able to make volumetric
determinations; this was not required in
our previous study. We also used differ-
ent segmentation techniques for the cur-
rent study, since several nodules had
complex relations with nearby structures
Small Pulmonary Nodules: Volumetrically Determined Growth Rates
and we found that k-means algorithms
did not provide sufficiently accurate seg-
mentation.
In evaluating the results of this prelim-
inary investigation, we found that, for
most nodules, the doubling time based
on either the area or the volume pro-
duced results that would have led to the
same conclusion. There were, however,
several striking exceptions that were in-
formative in leading to an understanding
of some of the reasons for these discrep-
ancies. Patient 1 (Fig 3) had a benign
doubling time based on 2D measure-
ments, yet had a malignant doubling
time based on 3D techniques. After anal-
ysis of the 3D image, the reason for this is
apparent. The growth of this nodule was,
in fact, asymmetric. It did not substan-
tially increase in size in the x-y plane;
rather, it grew along its z axis. Therefore,
even though the single image with the
tially. Such growth cannot be detected
easily by using 2D techniques. Further-
more, patient 1 helped to confirm our
belief that tumors do not necessarily
grow symmetrically and that different
portions may grow at different rates and
clearly showed the advantage of volu-
metric techniques.
All of the malignant nodules had volu-
metric doubling times of 177 days or less.
The doubling times of some malignant
nodules, particularly those associated
with well-differentiated adenocarcinoma,
can, however, be much slower. Benign
nodules either decreased in size, which re-
sulted in a negative volumetric doubling
time, or had volumetric doubling times
of 396 days or greater.
In patients 9 and 11, who had benign
nodules, marked differences between the
2D and 3D doubling times were found. In
both of these patients, benign disease was
expected on the basis of the change in 3D
volume, whereas malignancy would be
expected on the basis of the 2D area
change. Careful evaluation of the image
data in these studies showed that there
was greater difficulty in segmenting these
nodules, particularly on the CT images
containing the nodule area with the larg-
est diameter. Again, this points to the
advantage of the 3D technique because it
does not rely on a single CT image for
making a determination. The potential
effects of any segmenting difficulty would
be averaged over the entire nodule volume,
thus minimizing the overall effect.
At this time, we cannot be certain the
extent to which the differences seen in
doubling times based on the different
255
techniques (volume vs area) are due to
differences in the growth patterns of the
nodules themselves or differences in the
measuring techniques. We have given
examples in which one or the other was
more likely. In patients 2, 5, and 13, the
similarity of the results is striking. As we
accumulate additional cases and further
refine our segmentation techniques, we
presumably will be able to gain addi-
tional insight. It is clear, however, on the
basis of synthetic nodule studies, that CT
is capable of helping make these determi-
nations, assuming that segmentation tech-
niques for real nodules approach the re-
liability they do for synthetic nodules.
Although we had no overlap between
benign and malignant doubling times
based on 3D measurements, slow-grow-
ing malignant nodules may have dou-
bling times suggestive of benign disease.
There is even evidence to suggest that
these slower growing malignant nodules
behave less aggressively than the rapidly
growing ones (18). However, as we gain
experience, we will be able to develop a
scheduling function that will allow us to
decide when to perform additional repeat
studies to ascertain growth and to better
understand the probability that a partic-
ular growth rate implies either malig-
nancy or benignity. Prospective evalua-
tion of the techniques outlined in this
article should be performed toward the
refinement of such a model, as well as the
overall method. It seems likely that for
certain doubling times, further diagnos-
tic studies will be recommended.
At this time, we still need to learn more
about the growth of these small nodules,
as well as our measuring techniques, be-
256 Radiology October 2000
cause this preliminary study was limited
by the relatively small number of cases
evaluated. It is also important to note,
however, that we focused on pulmonary
nodules smaller than 10 mm, because
other diagnostic techniques (eg, biopsy)
are performed more easily in nodules
larger than 10 mm and it is more difficult
to justify periods of observation in these
cases to establish growth rates. As with all
other diagnostic tests, the accuracy of
volumetric measurements declines with
decreasing nodule size, primarily due to
the increasing partial voxel averaging.
Despite this, 3D volumetric techniques
appear to be a promising new tool for
obtaining additional information about
smaller nodules.
References
1. Friese SA, Rieber A, Fleiter T, Brambs HJ,
Claussen CD. Pulmonary nodules in spi-
ral volumetric and single-slice computed
tomography. Eur J Radiol 1994; 18:48 51.
2. Diederich S, Lenzen H, Windmann R, et
al. Pulmonary nodules: experimental and
clinical studies at low-dose CT. Radiology
1999; 213:289 298.
3. Chiou YS, Lure YM. Hybrid lung nodule
detection (HLND) system. Cancer Lett
1994; 77:119 126.
4. Zhu X, Lee KN, Levin DL, et al. Temporal
image database design for outcome anal-
ysis of lung nodule. Comput Med Imag-
ing Graph 1996; 20:347356.
5. Kanazawa K, Kawata Y, Niki N, et al.
Computer-aided diagnosis for pulmonary
nodules based on helical CT images.
Comput Med Imaging Graph 1998; 22:
157167.
6. Yankelevitz DF, Gupta R, Zhao B, Hensch-
ke CI. Small pulmonary nodules: evalua-
tion with repeat CTpreliminary experi-
ence. Radiology 1999; 212:561566.
7. Armato SG III, Giger ML, Moran CJ,
Blackburn JT, Doi K, MacMahon H. Com-
puterized detection of pulmonary nod-
ules on CT scans. RadioGraphics 1999;
19:13031311.
8. Henschke CI, Miettinen OS, Yankelevitz
DF, Libby D, Smith JP. Radiographic
screening for cancer: new paradigm for its
scientific basis. Clin Imaging 1994; 18:
16 20.
9. Henschke CI, McCauley DI, Yankelevitz
DF, et al. Early Lung Cancer Action
Project: overall design and findings from
baseline screening. Lancet 1999; 354:99
105.
10. Munden RF, Pugatch RD, Liptay MJ, Sug-
arbaker DJ, Le LU. Small pulmonary le-
sions detected at CT: clinical importance.
Radiology 1997; 202:105110.
11. Lillington GA. Management of solitary
pulmonary nodules. Postgrad Med 1997;
101:145150.
12. Lillington GA. Management of the soli-
tary pulmonary nodule. Disease-a-month
1991; 37:271318.
13. Yankelevitz DF, Henschke CI. Does 2-year
stability imply that pulmonary nodules
are benign? AJR Am J Roentgenol 1997;
168:325328.
14. Jain AK. Fundamentals of digital image
processing. Englewood Cliffs, NJ: Prentice
Hall, 1989.
15. Harris KM, Adams H, Lloyd DC, Harvey
DJ. The effect on apparent size of simu-
lated pulmonary nodules of using three
standard CT window settings. Clin Radiol
1993; 47:241244.
16. Mizuno T, Masaoka A, Ichimura H, Shi-
bata K, Tanaka H, Niwa H. Comparison of
actual survivorship after treatment with
survivorship predicted by actual tumor-
volume doubling time from tumor diam-
eter at first observation. Cancer 1984; 53:
2716 2720.
17. Russ JC. The image processing handbook.
2nd ed. Boca Raton, Fla: CRC, 1995.
18. Spratt JS, Meyer JS, Spratt JA. Rates of
growth of human solid neoplasms: part I.
J Surg Oncol 1995; 60:137146.
Yankelevitz et al