Diffusion-Weighted and Dynamic Contrast-Enhanced MRI of Prostate Cancer: Correlation of Quantitative

Genitourinar y Imaging • Original Research
Oto et al.
DWI and DCE-MRI of Prostate Cancer
Genitourinary Imaging
Original Research
Diffusion-Weighted and Dynamic
Contrast-Enhanced MRI of Prostate
Cancer: Correlation of Quantitative
MR Parameters With Gleason
Score and Tumor Angiogenesis
Aytekin Oto1
Cheng Yang1
Arda Kayhan1
Maria Tretiakova2
Tatjana Antic2
Christine Schmid-Tannwald1
Scott Eggener 3
Gregory S. Karczmar 1
Walter M. Stadler4
Oto A, Yang C, Kayhan A, et al.
Keywords: angiogenesis, diffusion-weighted MRI,
dynamic contrast-enhanced MRI, Gleason score,
prostate cancer
Received March 15, 2011; accepted after revision
May 6, 2011.
Presented at the 2010 annual meeting of the Radiological
Society of North America, Chicago, IL.
Supported by the Illinois division of the American
Cancer Society.
Department of Radiology, University of Chicago,
5841 S Maryland Ave, MC 2026, Chicago, IL 60637.
Address correspondence to A. Oto
([email protected]).
Department of Pathology, University of Chicago,
Chicago, IL.
Department of Surgery, Section of Urology, University of
Chicago, Chicago, IL.
Department of Medicine, Section of Hematology/
Oncology, University of Chicago, Chicago, IL.
AJR 2011; 197:1382–1390
© American Roentgen Ray Society
1382 OBJECTIVE. The objective of our study was to investigate whether quantitative parameters derived from diffusion-weighted imaging (DWI) and dynamic contrast-enhanced MRI
(DCE-MRI) correlate with Gleason score and angiogenesis of prostate cancer.
MATERIALS AND METHODS. Seventy-three patients who underwent preoperative
MRI and radical prostatectomy were included in our study. A radiologist and pathologist located the dominant tumor on the MR images based on histopathologic correlation. For each
dominant tumor, the apparent diffusion coefficient (ADC) value and quantitative DCE-MRI
parameters (i.e., contrast agent transfer rate between blood and tissue [Ktrans], extravascular
extracellular fractional volume [ve], contrast agent backflux rate constant [kep], and blood
plasma fractional volume on a voxel-by-voxel basis [vp]) were calculated and the Gleason
score was recorded. The mean blood vessel count, mean vessel area fraction, and vascular
endothelial growth factor (VEGF) expression of the dominant tumor were determined using
CD31, CD34, and VEGF antibody stains. Spearman correlation analysis between MR and
histopathologic parameters was conducted.
RESULTS. The mean tumor diameter was 15.2 mm (range, 5–28 mm). Of the 73 prostate
cancer tumors, five (6.8%) had a Gleason score of 6, 46 (63%) had a Gleason score of 7, and
22 (30.1%) had a Gleason score of greater than 7. ADC values showed a moderate negative
correlation with Gleason score (r = –0.376, p = 0.001) but did not correlate with tumor angiogenesis parameters. Quantitative DCE-MRI parameters did not show a significant correlation
with Gleason score or VEGF expression (p > 0.05). Mean blood vessel count and mean vessel area fraction parameters estimated from prostate cancer positively correlated with kep (r =
0.440 and 0.453, respectively; p = 0.001 for both).
CONCLUSION. There is a moderate correlation between ADC values and Gleason
score and between kep and microvessel density of prostate cancer. Although the strength of
the correlations is insufficient for immediate diagnostic utility, these results warrant further
investigation on the potential of multiparametric MRI to facilitate noninvasive assessment of
prostate cancer aggressiveness and angiogenesis.
rostate cancer is the second leading cause of cancer-related
deaths in American men [1].
Since the introduction of prostate-specific antigen (PSA) testing for prostate cancer screening, the lifetime risk of being diagnosed with prostate cancer has
doubled from 9% to 18%, but nearly 50% of
these cancers have a low risk of progression
and will not lead to mortality [1–3]. Unfortunately, reliable noninvasive methods to differentiate significant from insignificant cancers are lacking and most patients with
localized prostate cancer are treated with
radical treatment methods, which can be associated with serious adverse effects [2]. The
most important predictors of prognosis in
prostate cancer are Gleason score and tumor
staging [4, 5]. Tumor-associated angiogenesis, measured as microvessel density (MVD),
can also provide prognostic information in a
variety of solid tumors including prostate
cancer and can serve as a prognostic indicator of recurrence and metastatic disease [6–
8]. However, accurate estimation of these parameters is possible only after radical
prostatectomy. Development of a noninvasive tool that could assess the biologic aggressiveness of prostate cancer at the time of
diagnosis would be a major advance and
would have a significant impact on the choice
of treatment.
AJR:197, December 2011
DWI and DCE-MRI of Prostate Cancer
Despite its limitations, multiparametric MRI including T2-weighted imaging,
MR spectroscopy, diffusion-weighted imaging (DWI), and dynamic contrast-enhanced
MRI (DCE-MRI) is currently the best imaging modality for the diagnosis and staging
of prostate cancer. One of the limitations of
conventional T2-weighted images is their inability to assess tumor aggressiveness. Only
one recent study has shown a negative correlation between Gleason scores and tumormuscle signal intensity ratios on T2-weighted
images [9]. More recently applied functional MRI sequences such as DWI and DCEMRI have the potential to provide information about the tumor microenvironment and
angiogenesis and, hence, about the biologic
aggressiveness of the tumor [10–14].
In this study, our purpose was to evaluate the potential of MRI to predict histologic prognostic parameters by investigating the
correlation between quantitative DCE-MRI
parameters (calculated using a two-compartment model) and the apparent diffusion coefficient (ADC) with histologic parameters
including Gleason score, vascular endothelial growth factor (VEGF), and MVD established from surgically resected specimens of
prostate cancer.
Materials and Methods
Study Patients
This retrospective study was conducted with
an institutional review board–approved waiver
of informed consent and was in compliance with
HIPAA. Seventy-three consecutive prostate cancer
patients (median serum PSA level, 5.5 ng/mL;
range of PSA levels, 1.1–65.0 ng/mL; average age,
60.8 years; age range, 47–75 years) who underwent
endorectal MRI followed by radical prostatectomy
between September 2007 and December 2008
were identified from our institutional imaging
database. The average period between MRI and
prostatectomy was 44 days (range, 7–118 days).
MRI Protocols
All MR examinations were performed using
a 1.5-T scanner (Excite HD, GE Healthcare [n =
61]; or Achieva, Philips Healthcare [n = 12]). An
endorectal coil combined with a phased-array
surface coil was used for all examinations except
DCE-MRI examinations performed on the GE
scanner; for those examinations, only a phased-array
coil was used.
Immediately before the MR examination, 1
mg of glucagon was injected intramuscularly. We
imaged the entire prostate and oriented axial images
to be perpendicular to the rectal wall, guided by
sagittal images. A parallel imaging factor of 2 was
used in all sequences. The following axial, coronal,
and sagittal images were obtained: T2-weighted
fast spin-echo (FSE) (slice thickness, 3 mm), axial
T1 FSE, axial free-breathing DWI (b = 0, 1000,
and 1500 s/mm2), and axial free-breathing DCEMRI. Acquisition of DCE-MR images of the entire
prostate started 30 seconds before IV administration
of 0.1 mmol/kg of gadodiamide (Omniscan, GE
Healthcare) followed by a 20-mL saline flush at a
rate of 2.0 mL/s. Detailed acquisition protocols are
given in Appendix 1.
MRI-Histopathology Correlation
Surgical specimens of the entire prostate
after radical prostatectomy were inked and fixed
in 10% neutral buffered formalin for at least 24
hours. After dehydration, each specimen was cut
serially into 3-mm-thick sections from the apex
to base in transverse planes. Each serial section
was then either halved or quartered depending on
its size and was put in a cassette for processing,
paraffin embedding, and microtome cutting.
A genitourinary pathologist with more than
6 years’ experience in genitourinary pathology
at the time of the study reviewed the H and E
sections of the 73 prostate cancer patients and
using a four-quadrant approach (i.e., right anterior,
right posterior, left anterior, and left posterior)
recorded the size, location, and Gleason score of
each carcinoma on a schematic prostate diagram.
A radiologist who had 8 years of experience
in prostate MRI determined the locations of
the carcinoma on T2-weighted images on the
basis of these diagrams and in consultation
with the pathologist. Each H and E section was
then visually matched to a corresponding T2weighted image on the basis of the location of the
ejaculatory ducts, the dimension of the prostate,
any identifiable benign prostatic hyperplasia
nodule, and the approximate distance from the
base or apex. To be considered a match, a focus
of carcinoma must be in the same anterior or
posterior half of the prostate and must be at the
same superior-to-inferior level of the prostate.
For the portions of tumor foci that were invisible
on the matched T2-weighted image, the locations of
the tumor were determined by its position relative
to landmarks, such as the ejaculatory ducts, and its
anteroposterior position. The largest tumor focus
that could be matched confidently to a T2-weighted
MR image by consensus of the radiologist and
pathologist was included for analysis in each patient.
Regions of interest (ROIs) of prostate cancer were
then drawn manually on T2-weighted images and
subsequently determined on other MR images with
the help of image-registration software in our PACS
system (iSite Radiology System, Philips Healthcare).
4-µm sections of paraffin-embedded tissue
were stained with H and E and were bound to a
secondary antibody (EnvisionTM+ Kit, cat ##
K4001 and K4007, Dako) using a horseradish
peroxidase–labeled biotin-free dextrose-based
polymer complex. In brief, paraffin sections were
deparaffinized in xylenes, rehydrated through the
addition of graded ethanol solutions to distilled
water, and then washed in Tris-buffered saline.
Antigens were retrieved using either a hightemperature treatment in citrate buffer for CD31
and VEGF for 15 minutes in a microwave oven
or proteinase K digestion for CD34 (0.04 mg/mL)
for 5 minutes at 37°C. Endogenous peroxidase
activity was quenched by incubation in 3% H2O2 in
methanol for 5 minutes. Nonspecific binding sites
were blocked using serum-free blocking solution
(Protein Block, #X0909, Dako) for 20 minutes.
Tissue sections were then incubated for 1 hour
at room temperature with the mouse monoclonal
antibodies against CD31 (clone JC70A, 1:40,
Dako) and CD34 (clone QBEnd, 1:50, Novocastra)
and rabbit polyclonal antibody against VEGF
(sc-152, 1:100). This step was followed by 30
minutes of incubation with either goat antimouse
or antirabbit IgG conjugated to an horseradish
peroxidase–labeled polymer (EnvisionTM+
System, Dako). Slides were then developed for 5
minutes with 3-3′-diaminobenzidine chromogen,
counterstained with hematoxylin, and covered
with a cover slip. Negative control experiments
were performed by substituting the primary
antibody step with nonimmune mouse or rabbit
Automated Image Analysis
An automated MVD count was performed as
described earlier using an automated system
(Automated Cellular Imaging System [ACIS],
version II, ChromaVision Medical Systems) [15].
The ACIS automatically loads conventional
immunohistochemistry microscope slides for
scanning at high resolution on a bright field
microscope (Olympus BX45, Olympus) using a
digital camera (CCD camera, Sony). Then a
computer program creates a histologic re­
struction of an entire section image composed of
all individual microscopic fields. The image
analysis system’s MVD application was
configured for vessel detection using CD31- and
CD34-stained slides. This configuration was
based on applying chromogen masks for high
chromogenic staining (brown threshold) and
hematoxylin counterstaining (blue threshold) and
the minimal and maximal sizes of the vessels. In
each section, 10 representative nonoverlapping
AJR:197, December 20111383
Oto et al.
areas of dominant cancer were selected. Each
selected area was equal to 100× microscope FOV
in diameter (313,841 µm 2) and included at least
2000 cells.
In each selected area the following mea­sure­
ments were captured digitally: the mean vessel
count per area; the mean vessel area (MVA),
which is the square microns occupied by positively
stained vessels; and the MVA fraction, which is the
vessel density calculated as ratio of MVA in the
total area of counterstained tissue. The parameter
used to evaluate the VEGF expression was based
on the area and intensity of the brown stain, called
“integrated optical density,” or “IOD,” per 10 µm2.
MRI Analysis
Diffusion-weighted imaging analysis—Apparent
diffusion coefficient (ADC) maps were generated
from diffusion-weighted (DW) images with
commercial diffusion-analysis software (Advantage
Windows, version 4.2.3, GE Healthcare; or extended
MR work space, version, Philips Healthcare).
Because of the significantly lower spatial resolution
of DW images than T2-weighted images, simply
converting the ROIs drawn on T2-weighted images
into ROIs on DW images using image-registration
software would result in significant partial volume
effect on the margin of the ROIs. To remedy this
problem, two radiologists, who had 8 years and 1
year of experience with prostate MRI, manually
drew ROIs on DW images independently using the
ROIs on T2-weighted images as a reference and with
guidance of automatic image-registration software.
Based on the image coordinates information
recorded in the DICOM head files of T2-weighted
images and DW images, the imaging-registration
software enabled us to pinpoint the tumor region on
DW images corresponding to the tumor ROI selected
on the T2-weighted images. In drawing the ROIs, the
DW images were magnified and the largest possible
oval-shaped ROI was placed on the area of interest
excluding the blurred margin caused by lower spatial
resolution. Each radiologist drew one ROI per each
focus. The mean of the two ADC measurements was
accepted as the ADC of the tumor focus.
DCE-MRI analysis—We used a previously
published approximation [16] to convert the
measured signal S(t), where S is signal and t is
time, from T1-weighted DCE-MRI data acquired
with a small TR/TE, median flip angle, and
standard contrast agent dosage into contrast agent
concentration (Ct):
where r1 is the relaxivity coefficient (~ 4.5 1/s
mM at normal body temperature and 1.5 T for
Omniscan) and T1 is T1 relaxation time. The
conversion coefficient in the first set of brackets is
usually calculated using a T1 value reported in the
literature and a baseline signal, both of a reference
tissue (RT). This method basically uses normal
tissue as the reference to infer the unenhanced T1
value of other regions, which is then used to
calculate contrast agent concentration. In this
study, we used normal prostate tissue as the
reference tissue and used the T1 value of 1317 ± 85
ms [17].
The Tofts model [18] of time dependence of
contrast agent concentration Ct (t) was applied
to calculate the contrast agent transfer rate
between blood and tissue, Ktrans ; the extravascular
extracellular fractional volume, νe; and the blood
plasma fractional volume, νp, on a voxel-by-voxel
basis. The contrast agent backflux rate constant
kep was also calculated as follows:
kep = Ktrans / νe.
Average parameters were then calculated within
each ROI.
When applying the Tofts model, for contrast
agent arterial input function (AIF) we used the
average AIFs estimated for the GE Healthcare
and Philips Healthcare scanners based on a
previous clinical DCE-MRI study that used a
multiple reference tissue method [19]. It has been
shown previously that a realistic population AIF
value can produce reproducible estimates of the
kinetic parameters from DCE-MRI data [20]
that correlate strongly with reference standard
estimates from dynamic CT data [19]. The tumor
ROI on DCE-MR images was automatically
converted from the ROI on T2-weighted images
using image-registration software, and the
average DCE-MRI parameters were calculated
for each tumor ROI.
Statistical Analysis
Statistical analysis was performed using
statistics software (SPSS, version 17, SPSS)
for Microsoft Windows. Bivariate plots were
generated and analyses were conducted using the
Spearman rank correlation coefficient to evaluate
the associations of quantitative MRI parameters
with commonly used predictors and prognostic
markers including the Gleason score, MVD, and
VEGF expression. The significance level was
set at 0.00167 = 0.05 / 30 based on Bonferroni
correction, where 30 was the total number of
correlation tests we explored (Table 1). Internal
consistency for ADC measurements was assessed
using Cronbach’s alpha statistic.
TABLE 1: Spearman Rank Correlation Coefficients Between Histopathologic Parameters and Quantitative DiffusionWeighted Imaging and Dynamic Contrast-Enhanced MRI Parameters
Gleason score
Mean vessel counts per areab
Mean microvessel area fraction
Mean vessel counts per areab
Mean microvessel area fraction
Note—The significance level was set at 0.05 / 30 = 0.00167 based on Bonferroni correction. Statistically significant correlation coefficients are highlighted in boldface.
ADC = apparent diffusion coefficient, Ktrans = contrast agent transfer rate between blood and tissue, ve = extravascular extracellular fractional volume, kep = contrast
agent backflux rate constant, vp = blood plasma fractional volume on a voxel-by-voxel basis, VEGF = vascular endothelial growth factor.
aNumber of samples.
bArea = 313,841 µm2 (≥ 200 cells).
AJR:197, December 2011
DWI and DCE-MRI of Prostate Cancer
Seventy-three cancer foci, 64 peripheral zone (PZ) and nine transition zone (TZ)
cancer, with an average greatest dimension
of 15.2 mm (range, 5–28 mm) were included
in the analysis. A Gleason score of 6 was assigned in five cases (6.8%), a Gleason score
of 7 in 46 (63.0%), a Gleason score of 8 in
13 (17.8%), and a Gleason score of 9 in nine
(12.3%). Of 73 patients, 11 patients underwent prostatectomy at another institution so
histology slides were available for review but
tissue samples were not available for immunohistochemical staining. We could not ob-
tain quantitative DCE-MRI parameters in
seven patients: High-temporal-resolution
DCE-MRI data were not found in the data
archive because of impaired renal function
in four patients and there were excessive motion artifacts in another three patients. DW
images were not available in three patients to
calculate ADC.
Figure 1 shows the MR data and the H and
E– and immunohistochemistry-stained prostate samples from a representative 53-yearold patient. The MRI parameters and immunohistochemical parameters are described in
Table 2. Table 1 shows the Spearman rank
correlation coefficients of the histopathologic parameters and quantitative DWI and
DCE-MRI parameters for comparison.
Correlation Between ADC Values and
Gleason Scores
Internal consistency of ADC measurements for both readers was excellent (Cronbach’s alpha = 0.982). Gleason scores had a
statistically significant and moderate negative correlation with ADC measurements
(r = –0.376, p = 0.001). There was substantial overlap between the ADC values of tumors with different Gleason scores; however,
Fig. 1—53-year-old man with prostate cancer
(Gleason score for dominant tumor = 7).
A, Photomicrograph of H and E–stained prostate
sample shows prostate cancer in right peripheral
zone (arrow).
B–D, Axial T2-weighted MR image (B), apparent
diffusion coefficient map derived from diffusionweighted imaging (C), and map of contrast agent
transfer rate between blood and tissue (Ktrans ) (D)
derived from dynamic contrast-enhanced MRI (DCEMRI) show similar anatomic area. Dominant tumor is
indicated by arrows. Tumor region of interest on DCEMRI is shown by white contour in D.
E–G, Photomicrographs of serial histologic sections
from dominant prostate cancer tumor nodule shown
in A–D immunostained with CD31 (E), CD34 (F), and
vascular endothelial growth factor (G). In E and F,
vascular endothelium is stained dark brown. Original
magnification was ×100.
AJR:197, December 20111385
Oto et al.
TABLE 2: Descriptive Statistics for Immunohistochemical Parameters and MRI Parameters
No. of Samples
Mean vessel counts per areab
Mean microvessel area fraction
Mean vessel counts per areab
Mean microvessel area fraction
ADC (10 −3 mm2 /s)
Ktrans (min−1)
kep (min−1)
Note—VEGF = vascular endothelial growth factor, ADC = apparent diffusion coefficient, Ktrans = contrast agent transfer rate between blood and tissue, ve = extravascular
extracellular fractional volume, kep = contrast agent backflux rate constant, vp = blood plasma fractional volume on a voxel-by-voxel basis.
aVEGF expression was based on the area and intensity of the brown stain, called “integrated optical density,” or “IOD,” per 10 µm2 .
bArea = 313,841 µm2 (≥ 200 cells).
ADC values of tumors with a Gleason score
of 6 and those of tumors with a Gleason
score of 9 were relatively well separated. The
scatterplot of ADC values versus the Gleason scores of 70 dominant tumors in Figure
2 shows the trend that ADC values decreased
as Gleason scores increased.
Correlation Between Dynamic
Contrast-Enhanced MRI Parameters and
Histologic Findings
No significant correlation was observed
between any of the DCE-MRI parameters
and Gleason score. The mean blood vessel
count and mean vessel area fraction parameters calculated by CD31 staining positively
ADC (10-3 mm2/s)
Gleason Score
Fig. 2—Scatterplot of apparent diffusion coefficient
(ADC) values versus Gleason scores for 70 prostate
tumors in study group. ADC values were negatively
correlated with Gleason scores. Spearman rank
correlation coefficient, r, was –0.376 (p = 0.001).
correlated with kep (r = 0.440 and 0.453, respectively; both, p = 0.001). We should note
that even though mean blood vessel count and
mean vessel area fraction parameters showed
a moderate negative correlation with ve (r =
–0.327 and –0.328; p = 0.015 and 0.014, respectively) and a moderate positive correlation with vp (r = 0.312 and 0.393; p = 0.020
and 0.003, respectively), these correlations
did not reach a statistical significance after
Bonferroni correction. VEGF expression and
MVD parameters calculated by CD34 staining had no significant correlations with any of
the MRI parameters.
Our results show that there is a moderate negative correlation between ADC values calculated from DW images and Gleason
score of prostate cancer obtained from radical prostatectomy specimen. The kep (i.e.,
contrast agent backflux rate constant) derived from DCE-MRI moderately correlates
with angiogenesis parameters (mean blood
vessel count and mean vessel area fraction),
but no correlation was found between any of
the quantitative DCE-MRI parameters and
the Gleason score or VEGF expression of the
prostate cancer.
The association between lower ADC values and higher Gleason scores has been previously described in studies correlating DW
images with transrectal ultrasound–guided
biopsy results [10, 11]. Tamada et al. [10] reported a correlation coefficient, r, between
ADC and Gleason score of –0.497 in PZ
cancer and of –0.343 in TZ cancer. More recently, Woodfield et al. [11] found statistically significant differences between the ADC
values of PZ tumors with Gleason scores of
6 and 7 and those with Gleason scores of 6
and 8. In another study performed in 44 patients with prostate cancer, significant differences in tumor ADC values were reported
between patients with low-risk disease and
those with higher-risk disease [21]. The major limitations of these studies were the unreliable Gleason score obtained from needle
biopsies and difficulties in accurate localization of the biopsied tumor on MRI.
Needle biopsy leads to underestimation of
Gleason score in approximately 25% of the
cases compared with Gleason score established from prostatectomy specimen because
of biopsy sampling error and tumor heterogeneity [22, 23]. To overcome this limitation,
we used prostatectomy specimens for imaging correlation and establishing Gleason
score. Our results are similar to the results
of Tamada et al. [10]. Mazaheri et al. [24]
also observed lower ADC values for prostate tumors with higher Gleason scores from
prostatectomy. The correlation coefficient we
found was –0.376 in 70 prostate tumors composed of 61 PZ and nine TZ tumors.
A lower ADC value with a higher Gleason
score is most likely because of the increased
cellularity of tumors with a higher Gleason
score. Gibbs et al. [25] reported a trend of
increasing cell density of the prostate cancer
with increasing Gleason score. Cell density
increased from a mean of 14.5% for tumors
AJR:197, December 2011
DWI and DCE-MRI of Prostate Cancer
with a Gleason score of 6 to 21.9% for those
with a Gleason score of 8 or greater [25].
A significant negative correlation between
ADC and cellular density, or percentage area
of nuclei and cytoplasm, and a positive correlation between ADC and percentage area
of luminal space have also been previously
reported for prostate cancer [26–28]. Interestingly, even though ve is the extravascular
extracellular space fractional volume (thus,
a marker of cellular density), we did not find
a significant correlation between Gleason
score and any of the DCE-MRI parameters
including ve. This may be due to the fact that
tissue architecture parameters such as nucleus-cytoplasm ratio and area of glandular luminal space can affect ADC measurements
whereas they are not included in the calculation of ve.
Microvascularity is considered an important marker for neoangiogenesis, which in
turn is responsible for local growth and metastasis in tumors [29]. MVD has been reported to be associated with Gleason score,
tumor staging, recurrence, metastatic potential, and patient outcome in prostate cancer
[6–8, 30–34]. However, somewhat contradictory results have also been reported regarding the association between MVD and
the biologic behavior of prostate cancer [35,
36]. These contradictory results may be attributed to the differences in the composition of the study groups and the method used
for quantification of MVD. The choice of
antibody to stain vessels (CD31 vs CD34),
the method of selection of the area for vessel count (selection of “hot spots,” the areas
with the highest MVD in the tumor, vs random sampling of entire tumor region), and
the actual counting method (manual vs automated counting methods) may all influence the MVD measurements [37]. In our
study, we used both CD31 and CD34 for immunohistochemical staining and performed
sampling of an entire tumor region with 10
randomly placed ROIs over each tumor for
MVD calculations rather than selectively
including the hot spots [6]. Localization of
these hot spots on histology slides is prone
to observer bias. By sampling the entire tumor region, we aimed to improve the reproducibility of our results by minimizing this
bias [37, 38]. We also used automated digital
quantification of MVD using ACIS. Application of automated digital image analysis has
been shown to enhance reproducibility of
both the selection of the measurement area
and the actual vessel counts [38, 39].
Limited studies evaluating the correlation
between DCE-MRI parameters and angiogenesis markers of prostate cancer have provided contradicting results [12, 13]. Schlem­
mer et al. [12] reported significant correlation
between contrast exchange rate constant
(k21) (also called kep) and MVD, whereas
Kiessling et al. [13] found no such correlation. Both studies used the Brix model for
DCE-MRI analysis but used different antibodies (Schlemmer et al., CD31; Kiessling et
al., CD34) to calculate MVD. In our study,
only the MVD measured by CD31 staining
showed moderate but significant correlation
with kep, similar to the results of Schlemmer
et al. Our results suggest that the conflicting
results of two previous studies may at least
partly be due to the different antibodies used
in MVD calculation. CD31 is the most sensitive pan-endothelial marker: It stains large
and small vessels with equal signal intensity
as well as blood vessels in normal tissue and
tumor tissue [7]. On the other hand, CD34
is known to stain perivascular stromal cells
and some lymphatic vessels, which may have
contributed to falsely elevated MVD [40]. In
prostate specimens, the widely used CD34
monoclonal antibody is suggested to be less
optimal for microvessel staining because of
abundant stromal staining [37].
More recently, Franiel et al. [14, 41] studied the correlation between histologic parameters and the quantitative parameters
obtained from their dynamic dual-contrastenhanced MR data of prostate cancer, chronic prostatitis, and normal prostate tissue
from 35 patients. Similar to our study, they
used CD31 antibody and randomly selected
ROIs for MVD calculation. However, they
used a more complex three-compartmental
model to analyze the DCE-MRI data [41].
They found poor correlation between blood
volume and MVD (r = 0.252) and no correlation between blood volume and MVA or between interstitial volume measured by MRI
and histologic mean interstitial area.
In our study, correlations between quantitative perfusion parameters and histologic
parameters were moderate at best and only a
correlation between kep and MVD assessed by
CD31 reached a statistical significance. The
ve parameter had a moderate negative correlation and Ktrans had a very weak positive correlation with MVD. The kep is a composite parameter, equal to Ktrans / ve. The compounding
effect of Ktrans and ve may be the main reason for better correlation of kep with MVD.
We also observed that, even though it did not
reach statistical significance, vp correlated to
MVD assessed by CD31 better than Ktrans.
These findings may be explained by the more
direct relation between vp, the blood plasma
fractional volume, and MVD, whereas Ktrans
can be affected by both the perfusion and permeability of the vessels in tumor, so its relation to the MVD is more distant.
VEGF is a potent cytokine that supports
development of tumor vessels and its expression in the tumor correlates with cancer prognosis [42]. Immunohistochemical studies
have shown that human prostate cancer cell
tissues stained positively for VEGF, whereas
benign prostate tissue displayed little VEGF
staining [43]. Studies using prostate cancer
models also implied an association between
tumor metastatic capacity and VEGF expression [44]. Its usefulness as a prognostic factor is strongly suggested but remains to be
clarified, despite the strong evidence indicating its involvement in the growth process of
prostate cancer [45]. In our study, we did not
find any significant correlation between the
DCE-MRI parameters and VEGF expression. This may be partly explained by the
complex nature of the angiogenesis regulation involving many angiogenic factors and
inhibitors in addition to VEGF.
Our study has several limitations. First,
it is a retrospective study with a relatively
small sample size that could have been influenced by selection and verification biases.
The numbers of tumors with Gleason scores
of 6 and 9 were also small. At our institution,
preoperative MRI is used as the standard of
care for high-risk patients and this protocol explains the relatively small percentage
of tumors with a Gleason score of 6 in this
retrospective cohort. Second, the spatial correlation of a lesion between MR images and
histologic sections carries inherent limitations. Third, the moderate correlation or the
lack of correlation between MRI and histopathologic parameters might also have contributed to the errors in their measurement.
For example, Gleason score was subjectively
graded and human factors such as experience
will surely play a role in the accuracy of its
reading. For our DCE-MRI data, we used a
population AIF from a previous study to estimate the quantitative parameters. This analysis method was robust and easy to apply but
is less ideal than using individualized AIFs
and it inevitably contributed to some errors
in the estimated quantitative parameters. The
lack of unenhanced T1 mapping is an additional limitation of this study. As a result, the
AJR:197, December 20111387
Oto et al.
method of contrast agent concentration calculation relied on using normal prostate as
the reference tissue, which inevitably introduced observer variability. Fourth, despite
the similar protocols used, MRI studies were
performed using different scanners made by
different manufacturers (GE Healthcare and
Philips Healthcare) with different sequence
parameters. All of our MR scanners are routinely calibrated as a part of quality assurance study; however, there may be variation
between the MRI parameters derived from
different scanners. Furthermore, there is evidence that at least for ADC parameters interscanner variation is reasonable (± 5%) [46].
In conclusion, our results showed that
there is a moderate correlation between ADC
values derived from DWI and Gleason score
and between kep and MVD of prostate cancer. Although the strength of the correlations
is insufficient for immediate diagnostic utility, these results warrant further investigation
on the potential of multiparametric MRI to
facilitate noninvasive assessment of prostate
cancer aggressiveness and angiogenesis.
We thank Chuanhong Liao for helpful discussion on the statistical analysis.
1.Jemal A, Siegel R, Xu J, Ward E. Cancer statistics.
CA Cancer J Clin 2010; 60:277–300
2.Klotz L. Active surveillance versus radical treatment for favorable-risk localized prostate cancer.
Curr Treat Options Oncol 2006; 7:355–362
3.Eggener SE, Mueller A, Berglund RK, et al. A
multi-institutional evaluation of active surveillance for low risk prostate cancer. J Urol 2009;
181:1635–1641; discussion, 1641
4.Gleason DF, Mellinger GT; Veterans Administration Cooperative Urological Research Group. Prediction of prognosis for prostatic adenocarcinoma
by combined histological grading and clinical
staging: 1974. J Urol 2002; 167:953–958; discussion, 959
5.Kattan MW, Scardino PT. Prediction of progression: nomograms of clinical utility. Clin Prostate
Cancer 2002; 1:90–96
6.Lissbrant IF, Stattin P, Damber JE, Bergh A. Vascular density is a predictor of cancer-specific survival in prostatic carcinoma. Prostate 1997;
7.de la Taille A, Katz AE, Bagiella E, et al. Microvessel density as a predictor of PSA recurrence
after radical prostatectomy: a comparison of
CD34 and CD31. Am J Clin Pathol 2000;
8.Weidner N, Carroll PR, Flax J, Blumenfeld W,
Folkman J. Tumor angiogenesis correlates with
metastasis in invasive prostate carcinoma. Am J
Pathol 1993; 143:401–409
9.Wang L, Mazaheri Y, Zhang J, Ishill NM, Kuroiwa K, Hricak H. Assessment of biologic aggressiveness of prostate cancer: correlation of MR
signal intensity with Gleason grade after radical
prostatectomy. Radiology 2008; 246:168–176
10.Tamada T, Sone T, Jo Y, et al. Apparent diffusion
coefficient values in peripheral and transition
zones of the prostate: comparison between normal and malignant prostatic tissues and correlation with histologic grade. J Magn Reson Imaging
2008; 28:720–726
11.Woodfield CA, Tung GA, Grand DJ, Pezzullo JA,
Machan JT, Renzulli JF 2nd. Diffusion-weighted
MRI of peripheral zone prostate cancer: comparison of tumor apparent diffusion coefficient with
Gleason score and percentage of tumor on core
biopsy. AJR 2010; 194:1013; [web]W316–W322
12.Schlemmer HP, Merkle J, Grobholz R, et al. Can
pre-operative contrast-enhanced dynamic MR
imaging for prostate cancer predict microvessel
density in prostatectomy specimens? Eur Radiol
2004; 14:309–317
13.Kiessling F, Lichy M, Grobholz R, et al. Simple
models improve the discrimination of prostate
cancers from the peripheral gland by T1-weighted
dynamic MRI. Eur Radiol 2004; 14:1793–1801
14.Franiel T, Ludemann L, Rudolph B, et al. Prostate
MR imaging: tissue characterization with pharmacokinetic volume and blood flow parameters
and correlation with histologic parameters. Radiology 2009; 252:101–108
15.Westerhoff M, Tretiakova M, Hovan L, Miller J,
Noffsinger A, Hart J. CD61, CD31, and CD34 improve diagnostic accuracy in gastric antral vascular ectasia and portal hypertensive gastropathy:
an immunohistochemical and digital morphometric study. Am J Surg Pathol 2010; 34:494–501
16.Medved M, Karczmar G, Yang C, et al. Semiquantitative analysis of dynamic contrast enhanced MRI in cancer patients: variability and
changes in tumor tissue over time. J Magn Reson
Imaging 2004; 20:122–128
17.de Bazelaire CMJ, Duhamel GD, Rofsky NM, Alsop DC. MR imaging relaxation times of abdominal and pelvic tissues measured in vivo at 3.0 T:
preliminary results. Radiology 2004; 230:652–
18.Tofts PS, Brix G, Buckley DL, et al. Estimating
kinetic parameters from dynamic contrast-enhanced T(1)-weighted MRI of a diffusable tracer:
standardized quantities and symbols. J Magn Reson Imaging 1999; 10:223–232
19.Yang C, Stadler WM, Karczmar GS, Milosevic
M, Yeung I, Haider MA. Comparison of quanti-
tative parameters in cervix cancer measured by
dynamic contrast-enhanced MRI and CT. Magn
Reson Med 2010; 63:1601–1609
20.Parker GJM, Roberts C, MacDonald A, et al. Experimentally-derived functional form for a population-averaged high-temporal-resolution arterial
input function for dynamic contrast-enhanced
MRI. Magn Reson Med 2006; 56:993–1000
21.deSouza NM, Riches SF, Vanas NJ, et al. Diffusion-weighted magnetic resonance imaging: a potential non-invasive marker of tumour aggressiveness in localized prostate cancer. Clin Radiol
2008; 63:774–782
22.Steinberg DM, Sauvageot J, Piantadosi S, Epstein
JI. Correlation of prostate needle biopsy and radical prostatectomy Gleason grade in academic and
community settings. Am J Surg Pathol 1997;
23.Freedland SJ, Kane CJ, Amling CL, et al. Upgrading and downgrading of prostate needle biopsy specimens: risk factors and clinical implications. Urology 2007; 69:495–499
24.Mazaheri Y, Hricak H, Fine SW, et al. Prostate
tumor volume measurement with combined T2weighted imaging and diffusion-weighted MR:
correlation with pathologic tumor volume. Radiology 2009; 252:449–457
25.Gibbs P, Liney GP, Pickles MD, Zelhof B, Rodrigues G, Turnbull LW. Correlation of ADC and
T2 measurements with cell density in prostate
cancer at 3.0 Tesla. Invest Radiol 2009; 44:572–
26.Wang XZ, Wang B, Gao ZQ, et al. Diffusionweighted imaging of prostate cancer: correlation
between apparent diffusion coefficient values and
tumor proliferation. J Magn Reson Imaging 2009;
27.Zelhof B, Pickles M, Liney G, et al. Correlation of
diffusion-weighted magnetic resonance data with
cellularity in prostate cancer. BJU Int 2009;
28.Langer DL, van der Kwast TH, Evans AJ, et al.
Prostate tissue composition and MR measurements: investigating the relationships between
ADC, T2, K-trans, v(e), and corresponding histologic features. Radiology 2010; 255:485–494
29.Folkman J. The role of angiogenesis in tumor
growth. Semin Cancer Biol 1992; 3:65–71
30.Brawer MK, Deering RE, Brown M, Preston SD,
Bigler SA. Predictors of pathologic stage in prostatic carcinoma: the role of neovascularity. Cancer 1994; 73:678–687
31.Salomao DR, Graham SD, Bostwick DG. Microvascular invasion in prostate cancer correlates
with pathologic stage. Arch Pathol Lab Med 1995;
32.Bettencourt MC, Bauer JJ, Sesterhenn IA, Connelly RR, Moul JW. CD34 immunohistochemical
AJR:197, December 2011
DWI and DCE-MRI of Prostate Cancer
assessment of angiogenesis as a prognostic marker for prostate cancer recurrence after radical
prostatectomy. J Urol 1998; 160:459–465
33.Mehta R, Kyshtoobayeva A, Kurosaki T, et al. Independent association of angiogenesis index with
outcome in prostate cancer. Clin Cancer Res
2001; 7:81–88
34.Borre M, Offersen BV, Nerstrom B, Overgaard J.
Microvessel density predicts survival in prostate
cancer patients subjected to watchful waiting. Br
J Cancer 1998; 78:940–944
35.Rubin MA, Buyyounouski M, Bagiella E, et al.
Microvessel density in prostate cancer: lack of
correlation with tumor grade, pathologic stage,
and clinical outcome. Urology 1999; 53:542–547
36.Gettman MT, Bergstralh EJ, Blute M, Zincke H,
Bostwick DG. Prediction of patient outcome in
pathologic stage T2 adenocarcinoma of the prostate: lack of significance for microvessel density
analysis. Urology 1998; 51:79–85
37.van Niekerk CG, van der Laak JAWM, Boerger
ME, et al. Computerized whole slide quantifica-
tion shows increased microvascular density in
pT2 prostate cancer as compared to normal prostate tissue. Prostate 2009; 69:62–69
38.van der Laak JAWM, Westphal JR, Schalkwijk
LJM, et al. An improved procedure to quantify
tumour vascularity using true colour image analysis: comparison with the manual hot-spot procedure in a human melanoma xenograft model. J
Pathol 1998; 184:136–143
39.Belien JAM, Somi S, de Jong JS, van Diest PJ,
Baak JPA. Fully automated microvessel counting
and hot spot selection by image processing of
whole tumour sections in invasive breast cancer. J
Clin Pathol 1999; 52:184–192
40.Lindenmayer AE, Miettinen M. Immunophenotypic features of uterine stromal cells: CD34 expression in endocervical stroma. Virchows Arch
1995; 426:457–460
41.Franiel T, Ludemann L, Rudolph B, et al. Evaluation of normal prostate tissue, chronic prostatitis,
and prostate cancer by quantitative perfusion
analysis using a dynamic contrast-enhanced in-
version-prepared dual-contrast gradient echo sequence. Invest Radiol 2008; 43:481–487
42.Kut C, MacGabhann F, Popel AS. Where is VEGF
in the body? A meta-analysis of VEGF distribution in cancer. Br J Cancer 2007; 97:978–985
43.Ferrer FA, Miller LJ, Andrawis RI, et al. Vascular
endothelial growth factor (VEGF) expression in
human prostate cancer: in situ and in vitro expression of VEGF by human prostate cancer cells. J
Urol 1997; 157:2329–2333
44.Melnyk O, Zimmerman M, Kim KJ, Shuman M.
Neutralizing anti-vascular endothelial growth
factor antibody inhibits further growth of established prostate cancer and metastases in a preclinical model. J Urol 1999; 161:960–963
45.Delongchamps NB, Peyromaure M, Dinh-Xuan
AT. Role of vascular endothelial growth factor in
prostate cancer. Urology 2006; 68:244–248
46.Ivancevic MK, Meyer CR, Galban CJ, et al. Simple, universal phantom for multi-center apparent
diffusion coefficient (ADC) measurement. (abstr)
Proc Intl Soc Mag Reson Med 2009; 17:4166
(Appendix follows on next page)
AJR:197, December 20111389
Oto et al.
APPENDIX 1: MRI Acquisition Protocols
Protocol for GE Healthcare Scanner
Array spatial sensitivity-encoding technique (parallel imaging)
factor of 2 was used in all sequences.
Protocol for Philips Healthcare Scanner
Effective sensitivity-encoding (parallel imaging) factor of 2 was
used in all sequences.
T2-Weighted Imaging Parameters
• TR range/TE range = 3200–3500/90–100
• Matrix size = 192 × 256
• Echo-train length = 19
• Number of signals acquired = 4
• Section thickness = 3 mm
• Intersection gap = 0 mm
• FOV = 14–16 cm
T2-Weighted Imaging Parameters
• Resolution = 0.8 × 0.8 × 3 mm
• TR range/TE = 4300–5000/120
• Matrix size = 204 × 256
• Echo-train length = 24
• Number of signals acquired = 4
• Section thickness = 3 mm
• Intersection gap = 0 mm
• FOV = 14–18 cm
Diffusion-Weighted Imaging Parameters
• TR range/TE range = 7000–8000/80–90
• Matrix size = 128 × 128–224
• b values = 0, 1000, and 1500 s/mm 2
• Number of signals acquired = 4
• Slice thickness = 4 mm
• Gap = 0 mm
• FOV = 14–18 cm
T1-weighted, 3D, gradient-echo, and free-breathing axial dynamic contrast-enhanced MR images covering the entire prostate were
acquired starting 30 seconds before the IV administration of gadodiamide (Omniscan, GE Healthcare) at a dose of 0.1 mmol/kg, followed by a 20-mL saline flush at a rate of 2.0 mL/s.
Dynamic Contrast-Enhanced MRI Parameters
• TR range/TE range = 3.5–3.9/1.6–1.9
• Matrix size = 160 × 256
• Flip angle = 10°
• Interpolated slice thickness = 3 mm with temporal resolution of
5–12 seconds for approximately 5–7 minutes
Approximately 30–50 sets of images were acquired to monitor the
time course of contrast agent uptake and clearance within the prostate. The entire scanning protocol including patient setup was performed in less than 1 hour in all patients.
Diffusion-Weighted Imaging Parameters
• TR range/TE range = 3800–4200/80–90
• Matrix size = 128 × 128
• b values = 0, 1000, and 1500 s/mm 2
• Number of signals acquired = 4
• Slice thickness = 4 mm
• Gap = 0 mm
• FOV = 14–18 cm
Dynamic Contrast-Enhanced MRI Parameters
• Three-dimensional fast-field echo
• TR/TE = 5.5/2.1
• Matrix size = 199 × 292
• Interpolated section thickness = 3 mm with a temporal resolution of 3–5 seconds for approximately 7–9 minutes
The dose and administration of IV gadolinium was similar to the
GE protocol. Approximately 70–100 sets of images were acquired
and the entire scanning protocol including patient setup was performed in less than 1 hour in all patients.
AJR:197, December 2011