Note: This copy is for your personal non-commercial use only.

To order presentation-ready
Original Research  n  Neuroradiology copies for distribution to your colleagues or clients, contact us at www.rsna.org/rsnarights.

Glioblastoma Treated with

Concurrent Radiation Therapy
and Temozolomide Chemotherapy:
Differentiation of True Progression from
Pseudoprogression with Quantitative
Dynamic Contrast-enhanced MR
Imaging1
Tae Jin Yun, MD
Purpose: To explore the role of dynamic contrast material–enhanced
Chul-Kee Park, MD magnetic resonance (MR) imaging in the differentiation of
Tae Min Kim, MD true progression from pseudoprogression in patients with
Se-Hoon Lee, MD glioblastoma on the basis of findings in entirely newly devel-
Ji-Hoon Kim, MD oped or enlarged enhancing lesions after concurrent radiation
Chul-Ho Sohn, MD therapy and chemotherapy with temozolomide and to evalu-
Sung-Hye Park, MD ate the diagnostic performance of the quantitative pharma-
cokinetic parameters obtained at dynamic contrast-enhanced
Il Han Kim, MD
MR imaging, such as the volume transfer constant (Ktrans), the
Seung Hong Choi, MD extravascular extracellular space per unit volume of tissue(ve),
and the blood plasma volume per unit volume of tissue(vp).

Materials and
Methods:
1 cokinetic parameters were compared between the true pro-
 From the Institute of Radiation Medicine, Seoul National
University Medical Research Center, Seoul, Republic of Ko-
rea (T.J.Y., J.H.K., C.H.S., S.H.C.); Department of Radiology
(T.J.Y., J.H.K., C.H.S., S.H.C.), Department of Neurosurgery
(C.K.P.), Department of Internal Medicine, Cancer Research Results:
Institute (T.M.K., S.H.L.), Department of Pathology (S.H.P.),
and Department of Radiation Oncology, Cancer Research
Institute (I.H.K.), Seoul National University Hospital, 28
Yongon-dong, Chongno-gu, Seoul 110-744, Republic of
Korea; Center for Nanoparticle Research, Institute for Basic
Science, Seoul, Republic of Korea (S.H.C.); and School
of Chemical and Biological Engineering, Seoul National
University, Seoul, Republic of Korea (S.H.C.). Received
November 14, 2013; revision requested December 18; revi-
sion received May 29, 2014; accepted July 24; final version Conclusion:
accepted August 26. Supported by Bayer Healthcare, the
Korea Health Care Technology R&D Projects, the Korean
Ministry for Health, Welfare, and Family Affairs (grant
HI13C0015), the Research Center Program of the Institute
for Basic Science, and the National Research Foundation of
Korea (grant NRF-2013R1A1A2008332). Address corre-
spondence to S.H.C. (e-mail: verocay@snuh.org).
This prospective study had institutional review board ap-
proval; written informed consent was obtained from all pa-
tients. Thirty-three patients with histopathologically proven
glioblastoma who had undergone concurrent radiation ther-
apy and chemotherapy with temozolomide were included.
Dynamic contrast-enhanced MR imaging–derived pharmaco-
kinetic parameters, including Ktrans, ve, and vp, were calculated
for newly developed or enlarged enhancing lesions. Pharma-
gression (n = 17) and pseudoprogression (n = 16) groups by
using unpaired t tests and then multivariable analysis.
The mean Ktrans and ve were higher in the true progression
group than in the pseudoprogression group (mean Ktrans, 0.44
min21 6 0.25 [standard deviation] and 0.23 min21 6 0.10 for
true progression and pseudoprogression groups, respectively,
P = .004; and mean ve, 1.26 6 0.78 and 0.75 6 0.49 for true
progression and pseudoprogression groups, respectively, P =
.034). Multivariable analysis showed that mean Ktrans was the
only independently differentiating variable (P = .004).
Dynamic contrast-enhanced MR imaging–derived pharma-
cokinetic parameters, including Ktrans and ve, in the entire
newly developed or enlarged enhancing lesion may be use-
ful objective diagnostic tools in the differentiation of true
progression from pseudoprogression in patients with glio-
blastoma who have undergone concurrent radiation ther-
apy and chemotherapy with temozolomide.

q
 RSNA, 2014 q
 RSNA, 2014

830 radiology.rsna.org  n  Radiology: Volume 274: Number 3—March 2015

NEURORADIOLOGY: Quantitative MR Imaging for Evaluating Progression in Glioblastoma Yun et al
G
lioblastoma multiforme (GBM) is
the most common primary malig-
nant type of brain tumor in adults.
Surgical tumor resection followed by
concurrent radiation therapy and che-
motherapy with temozolomide (TMZ)
is the current standard therapy for pa-
tients with GBM (1). However, despite
the multiple treatment approaches, the
prognosis for patients with GBM is still
dismal (2,3).
It has recently been observed that
radiologic deterioration occurs im-
mediately after the end of concurrent
radiation therapy and chemotherapy
with TMZ in patients with GBM, with
spontaneous improvement without
further treatment other than adjuvant
TMZ in 20%–30% of patients under-
going their first postradiation magnetic
resonance (MR) imaging examina-
tion (4–6). This phenomenon, termed
Advances in Knowledge
nn Dynamic contrast-enhanced MR
imaging pharmacokinetic param-
eters, including the volume trans-
fer constant (Ktrans) and the ex-
travascular extracellular space
per unit volume of tissue (ve),
can enable the discrimination of
true progression from pseudo-
progression in newly developed
or enlarged enhancing lesions
after concurrent radiation
therapy and chemotherapy with
temozolomide for glioblastoma.
nn The mean Ktrans was 0.44 min21
6 0.25 (standard deviation) for
true progression and 0.23 min21
6 0.10 for pseudoprogression (P
= .004), while the mean ve was
1.26 6 0.78 for true progression
and 0.75 6 0.49 for pseudopro-
gression (P = .034).
nn Among the pharmacokinetic pa-
rameters, mean Ktrans , with a
sensitivity of 59% and a speci-
ficity of 94%, was the most
promising parameter in the dif-
ferentiation of true progression
from pseudoprogression among
the useful Ktrans and ve values
derived from the dynamic con-
trast-enhanced MR imaging data.
Radiology: Volume 274: Number 3—March 2015  n  radiology.rsna.org 831
pseudoprogression, seems to result
from transiently increased permeability
of the tumor vasculature from irradia-
tion (4). However, because there is no
established finding at conventional MR
imaging for differentiating between true
progression and pseudoprogression,
the Response Assessment in Neuro-On-
cology (RANO) criteria for high-grade
gliomas were revised such that within
the 1st 12 weeks after completion of
radiation therapy, when pseudopro-
gression is most prevalent, progression
can be determined only if the majority
of the new enhancement is outside the
radiation field or if there is pathologic
confirmation of progressive disease (6).
Therefore, if the measurable enhancing
lesion is located within the radiation
field or if it is not confirmed with path-
ologic examination, the diagnosis of
pseudoprogression is limited by RANO
criteria to 12 weeks after the comple-
tion of concurrent radiation therapy
and chemotherapy.
Despite the recent effort to distin-
guish true progression from pseudo-
progression by using conventional and
advanced MR imaging techniques, in-
cluding contrast material–enhanced im-
aging, diffusion-weighted imaging, and
perfusion-weighted imaging (7–12),
differentiation between true progres-
sion and pseudopression seems to be a
challenge. Dynamic contrast-enhanced
MR imaging is a noninvasive imaging
technique that can be used to derive
quantitative parameters that reflect mi-
crocirculatory structure and function in
imaged tissues. Ktrans and ve are phar-
macokinetic parameters well known
for reflecting permeability. Ktrans is the
Implication for Patient Care
nn With dynamic contrast-enhanced
MR imaging, we can make early
decisions regarding newly en-
hancing lesions after concurrent
radiation therapy and chemo-
therapy with temozolomide in
patients with glioblastoma and
provide proper treatment plans
to clinicians, thus improving
patients’ prognoses and quality
of life.
volume transfer constant between the
plasma and extravascular extracellular
space. ve is the extravascular extracel-
lular space per unit volume of tissue
and is also known as the leakage space
(13,14). vp is the blood plasma volume
per unit volume of tissue, which may
be a marker of angiogenic activity in a
tumor (10). Because abnormalities in
these parameters, particularly blood
volume and vascular permeability, are
inherent to tumor neovasculature, such
information is of great potential value
to the practice of oncology.
To our knowledge, however, there
have been no previous reports con-
cerning the use of dynamic contrast-
enhanced MR imaging to differentiate
true progression from pseudoprogres-
sion in patients with GBM who had
undergone surgical resection followed
by concurrent radiation therapy and
chemotherapy with TMZ. The purpose
of the present study was to explore the
role of dynamic contrast-enhanced MR
imaging in the differentiation of true
progression from pseudoprogression
in entire newly developed or enlarged
enhancing lesions after concurrent
radiation therapy and chemotherapy
with TMZ in patients with GBM and to
Published online before print
10.1148/radiol.14132632  Content code:
Radiology 2015; 274:830–840
Abbreviations:
AUC = area under the receiver operating characteristic
curve
GBM = glioblastoma multiforme
Ktrans = volume transfer constant
RANO = Response Assessment in Neuro-Oncology
TMZ = temozolomide
ve = extravascular extracellular space per unit volume of
tissue
VOI = volume of interest
vp = blood plasma volume per unit volume of tissue
Author contributions:
Guarantors of integrity of entire study, T.J.Y., S.H.C.; study
concepts/study design or data acquisition or data analysis/
interpretation, all authors; manuscript drafting or manu-
script revision for important intellectual content, all authors;
manuscript final version approval, all authors; literature
research, T.J.Y., T.M.K., S.H.L., J.H.K., C.H.S., S.H.P., S.H.C.;
clinical studies, all authors; statistical analysis, T.J.Y.,
S.H.C.; and manuscript editing, all authors
Conflicts of interest are listed at the end of this article.
NEURORADIOLOGY: Quantitative MR Imaging for Evaluating Progression in Glioblastoma Yun et al
evaluate the diagnostic performance of
the quantitative pharmacokinetic pa-
rameters obtained at dynamic contrast-
enhanced MR imaging, such as Ktrans,
ve, and vp. Our hypotheses were that
(a) the pharmacokinetic parameters
would reflect the vascular microenvi-
ronments of the newly developed or en-
larged enhancing lesions after the com-
pletion of concurrent radiation therapy
and chemotherapy and (b) there would
be differences in the pharmacokinetic
parameters between true progression
and pseudoprogression in GBM treated
with concurrent radiation therapy and
chemotherapy.
Materials and Methods
The institutional review board of Seoul
National University Hospital approved
this prospective study. Written informed
consent was obtained from all patients.
Patient Selection
Seventy-one patients with a new di-
agnosis of GBM who had undergone
concurrent radiation therapy and che-
motherapy with TMZ and six cycles of
adjuvant TMZ after surgery or biopsy
between September 2011 and April
2013 were enrolled in this prospective
study.
The inclusion criteria were as fol-
lows: The patient (a) had been given
a histopathologic diagnosis of GBM
without oligodendroglial components
on the basis of the World Health Or-
ganization criteria; (b) had undergone
concurrent radiation therapy and che-
motherapy with TMZ and six cycles of
adjuvant TMZ after surgery or biopsy;
(c) had undergone baseline MR imag-
ing performed with contrast enhance-
ment within 24–48 hours after surgery
but before subsequent concurrent
radiation therapy and chemotherapy
with TMZ; (d) had undergone fol-
low-up MR imaging, including dynamic
contrast-enhanced MR imaging, within
2 months after the end of concurrent
radiation therapy and chemotherapy
with TMZ; (e) had a newly appeared
or enlarged enhancing lesion (defined
as a bidimensionally enhancing lesion
with two perpendicular diameters of
832 radiology.rsna.org  n Radiology: Volume 274: Number 3—March 2015
at least 10 mm on the MR images)
inside the radiation field (6); and (f)
underwent follow-up MR imaging per-
formed with contrast enhancement for
the confirmation of true progression
or pseudoprogression after six cycles
of adjuvant TMZ. We excluded 38 pa-
tients (54% of all patients) in whom no
newly appearing or enlarged enhanc-
ing lesion was detected.
As a result, a total of 33 patients
with GBM (22 men and 11 women; age
range, 28–82 years) were included and
were confirmed to have true progres-
sion (n = 17 [52%]) or pseudoprogres-
sion (n = 16 [48%]) after the end of
adjuvant TMZ according to the RANO
criteria (6).
Image Acquisition
For each patient, after the completion
of concurrent radiation therapy and
chemotherapy with TMZ, follow-up MR
imaging studies were performed by us-
ing a 3.0-T imaging unit with a 32-chan-
nel head coil (Verio; Siemens Health-
care Sector, Erlangen, Germany). The
MR imaging examination included T1-
weighted imaging (with reconstruction
of multiplanar transverse and coronal
images) performed before and after
contrast enhancement with a sagittal
three-dimensional magnetization-pre-
pared rapid acquisition gradient-echo
sequence and a transverse fluid-atten-
uated inversion recovery sequence, as
well as transverse T2-weighted imag-
ing performed with turbo spin-echo
sequences. We performed the T1-
weighted imaging with three-dimen-
sional magnetization-prepared rapid
acquisition gradient-echo sequences
by using the following parameters:
repetition time msec/echo time msec,
1500/1.9; flip angle, 9°; matrix, 2563
232; field of view, 220 3 250; section
thickness, 1 mm; and number of sig-
nals acquired, one. The parameters for
axial fluid-attenuated inversion recov-
ery imaging were as follows: 9000/97;
inversion time, 2500 msec; flip angle,
130°; matrix, 384 3 348; field of view,
199 3 220; and section thickness, 5
mm. The parameters for the transverse
T2-weighted sequence were as follows:
5160/91; flip angle, 130°; matrix, 640 3
510–580; field of view, 175–199 3 220;
section thickness, 5 mm; and number
of signals acquired, three. Contrast-
enhanced T1-weighted imaging was
performed after intravenous adminis-
tration of gadobutrol (Gadovist; Bayer
Schering Pharma, Berlin, Germany)
at a dose of 0.1 mmol per kilogram of
body weight.
Dynamic contrast-enhanced MR im-
aging was performed with three-dimen-
sional gradient-echo T1-weighted imag-
ing after the intravenous administration
of gadobutrol (0.1 mmol/kg) at a rate
of 4 mL/sec by using a power injector
(Spectris; Medrad, Pittsburgh, Pa). A
30-mL bolus injection of saline followed
at the same injection rate. For each sec-
tion, 40 images were acquired at inter-
vals equal to the repetition time. The
parameters were as follows: 2.8/1.0; flip
angle, 10°; matrix, 192 3 192; section
thickness, 3 mm; field of view, 240 3
240 mm; voxel size, 1.25 3 1.25 3 3
mm3; pixel bandwidth, 789 Hz; and total
acquisition time, 1 minute 30 seconds.
Image Analysis
Dynamic contrast-enhanced MR images
were processed by using the MR perfu-
sion analysis method (Nordic ICE and
TumorEX; NordicNeuroLab, Bergen,
Norway), in which three-dimensional
gradient-echo T1-weighted imaging was
used for structural imaging. Deconvo-
lution with the arterial input function
(AIF) was performed in the pharma-
cokinetic model. For each tumor, the
AIF was determined semiautomatically
in intracranial tumor-supplying arteries
near the tumor. To define an AIF, pre-
liminary AIF curves were derived in 5
pixels near the tumor, and two neuro-
radiologists (T.J.Y., with 12 years of ex-
perience in neuroradiology, and S.H.C.,
with 13 years of experience) indepen-
dently excluded inappropriate AIF
curve(s) derived in undesired pixel(s).
A final AIF curve was generated by us-
ing the cluster analysis technique.
On the basis of the two-compart-
ment pharmacokinetic model pro-
posed by Tofts and Kermode (15), the
perfusion analysis method was used
to calculate pharmacokinetic parame-
ters, including Ktrans, ve, and vp. The
NEURORADIOLOGY: Quantitative MR Imaging for Evaluating Progression in Glioblastoma Yun et al
baseline T1 was fixed at 1000 msec in
this study (16).
Pharmacokinetic parameters, in-
cluding Ktrans, ve, and vp, were measured
in each volume of interest (VOI). The
perfusion analysis method was used to
calculate the pharmacokinetic param-
eters and to present these parameters
as parametric maps. The coregistration
between the structural images and para-
metric maps (color overlay) was accom-
plished automatically by using mutual
information on the basis of an algorithm
that facilitated the search for an optimal
rigid transformation aligning the two
data sets (17,18). Two neuroradiologists
(T.J.Y. and S.H.C.) manually defined the
cubic form, including all the enhancing
portions of the lesion independently.
This was followed by automatic defini-
tion of the VOI in which the degree of
enhancement was over the threshold at
the perfusion analysis workstation. On
a pixel-by-pixel basis, all values of the
parameters were calculated in the VOIs
in an axial plane. The overall values for
each tumor were obtained by summing
the values from every plane. Finally, the
total parametric values from the entire
tumor volume were acquired and re-
corded for each tumor. Representative
images describing the image analysis
process are shown in Figure 1.
Statistical Analysis
Clinical characteristics were compared
between the progression and pseudo-
progression groups by using the Fisher
exact test or the x2 test for categoric
variables and the unpaired Student t
test for noncategoric data. We as-
sessed interobserver agreement by
using the interclass correlation coeffi-
cient, the coefficient of variation, and
Bland-Altman plots. The Kolmogorov-
Smirnov test was used to determine
whether values were normally distrib-
uted. Differences in permeability pa-
rameters between the true progression
and pseudoprogression groups were
compared by using the unpaired t test.
For the quantitative analysis, cu-
mulative histograms were obtained
for each pharmacokinetic parameter
in which the cumulative number of
observations in all of the bins up to
Radiology: Volume 274: Number 3—March 2015  n  radiology.rsna.org 833
the specified bin was mapped on the
y-axis as a percentage. In addition,
the area under the receiver operating
characteristic curve (AUC) was used
to evaluate the ability of each per-
meability parameter to differentiate
true progression from pseudoprogres-
sion. We calculated the AUC at each
percentile point from the 1st to the
50th percentile of the cumulative his-
tograms and then found the percen-
tile point that had the highest AUC
(the xth percentile point is the point
at which x percent of the voxel values
that form the histogram are found to
the left of the histogram).
Thereafter, a multivariable stepwise
logistic regression model was used to
determine the best predictors of dif-
ferentiation between true progression
and pseudoprogression (19). In ad-
dition, leave-one-out cross validation
and the McNemar test were performed
to validate the value of the best pre-
dictor pharmacokinetic parameter in
differentiating true progression from
pseudoprogression.
Statistical analysis was performed
with commercially available software
(SPSS, version 12.0 for Windows,
SPSS, Chicago, Ill; and MedCalc, ver-
sion 9.3.0.0, MedCalc Software, Mar-
iakerke, Belgium). P , .05 was consid-
ered to indicate a statistically significant
difference.
Results
We determined whether any of the
pretreatment clinical characteristics,
including age, histologic findings, Kar-
nofsky Performance Status score, and
radiation dose, were predictors of true
progression (n = 17 [52%]) or pseudo-
progression (n = 16 [48%]) after con-
current radiation therapy and chemo-
therapy with TMZ. None of the clinical
parameters was a significant predictor
of true progression or pseudoprogres-
sion (Table 1).
Interobserver Agreement for Dynamic
Contrast-enhanced MR Imaging
Pharmacokinetic Parameters
The interclass correlation coefficients
for mean Ktrans, ve, and vp were 0.83
(95% confidence interval [CI]: 0.66,
0.91), 0.96 (95% CI: 0.91, 0.98), and
0.76 (95% CI: 0.52, 0.88), respec-
tively, while the coefficients of vari-
ation of quantitative agreement for
these parameters were 35.8%, 21.0%,
and 38.4%, respectively. According to
Bland-Altman plot analysis, the interob-
server agreement for mean Ktrans was
better than those for both mean ve and
mean vp (Fig 2).
Differences in Pharmacokinetic
Parameters between True Progression
and Pseudoprogression Groups
The mean Ktrans was higher in the true
progression group than in the pseudo-
progression group (0.44 min21 6 0.25
[standard deviation] vs 0.23 min21 6
0.10, respectively; P = .004). We also
observed a higher mean ve in the true
progression group than in the pseudo-
progression group (1.26 6 0.78 vs 0.75
6 0.49, respectively; P = .034). Mean
vp was not significantly different be-
tween the true progression and pseudo-
progression groups (P . .05).
The box-and-whiskers graphs for
the pharmacokinetic parameters are
shown in Figure 3.
Diagnostic Performance of Pharmaco­
kinetic Parameters in Differentiation of
True Progression from Pseudoprogression
Table 2 summarizes the AUCs, sensi-
tivities, specificities, and positive and
negative predictive values for the dif-
ferentiation of true progression (n =
17 [52%]) from pseudoprogression
(n = 16 [48%]) by using Ktrans and ve,
which showed significant differences
between the two groups. For Ktrans, the
10th percentile value showed the high-
est AUC for distinguishing true pro-
gression from pseudoprogression. The
AUCs, sensitivities, specificities, and
positive and negative predictive values
for mean Ktrans and the 10th percentile
value were 0.756, 59%, 94%, 91%, and
68% and 0.801, 82%, 75%, 78%, and
80%, respectively. In addition, for ve,
the 5th percentile value showed the
highest AUC for distinguishing true
progression from pseudoprogression.
The AUCs, sensitivities, specificities,
positive and negative predictive values
NEURORADIOLOGY: Quantitative MR Imaging for Evaluating Progression in Glioblastoma Yun et al

Figure 1

834 radiology.rsna.org  n Radiology: Volume 274: Number 3—March 2015

NEURORADIOLOGY: Quantitative MR Imaging for Evaluating Progression in Glioblastoma Yun et al

for mean ve and the 5th percentile value representative cases of true progres- and ve (mean and 5th percentile), both
were 0.750, 88%, 56%, 68%, and 82% sion and pseudoprogression are shown of which showed significant differences
and 0.813, 76%, 88%, 87%, and 78%, in Figure 4. between the true progression and pseu-
respectively. However, no significant doprogression groups. Multivariable
difference was revealed between the Results of Multivariable Analysis and logistic regression analysis showed that
AUCs for the mean and 10th percentile Validation of the Best Pharmacokinetic mean Ktrans was the only independently
Ktrans values or among the mean and Predictor for the Differentiation of True differentiating variable (P = .004).
5th percentile ve values (P . .05). Progression from Pseudoprogression Other Ktrans and ve parameters were
The Ktrans, ve, and vp maps, histo- For the multivariable analysis, we in- close alternatives (19). With use of the
grams, and cumulative histograms for cluded Ktrans (mean and 10th percentile) mean Ktrans of 0.347 min21 as a thresh-
old, dynamic contrast-enhanced MR
Table 1 imaging had a sensitivity of 59% (10 of
17 true progressions) and a specificity
Clinical Characteristics of the Patients of 94% (15 of 16 pseudoprogressions).
True Progression Pseudoprogression Thus, the overall accuracy of mean Ktrans
Characteristic Total (n = 33) Group (n = 17) Group (n = 16) P Value was 76%, with correct classifications in
25 of the 33 patients.
Mean age (y) 54.6 58.9 50.1 .076* In terms of the mean value of mean
Diagnosis of grade IV GBM 33 17 15 .485† Ktrans, leave-one-out cross validation
Karnofsky Performance Scale score ..99† showed a sensitivity of 59% (10 of 17)
  ,70 9 5 4
and a specificity of 94% (15 of 16),
  70 24 12 12
which were identical to the primary
Surgery .144‡
results.
 Biopsy 7 5 2
  Subtotal resection 13 8 5
  Gross total resection 13 4 9 Discussion
Mean radiation dose (Gy) 55.3 54.0 56.6 .395*
In this prospective study assessing the
Note.—Unless otherwise specified, data are numbers of patients. ability of dynamic contrast-enhanced
* Calculated by using the unpaired Student t test. MR imaging to discriminate true pro-
†
Calculated by using the Fisher exact test. gression from pseudoprogression in
‡
Calculated by using the x2 test. patients with GBM who had undergone

Figure 2

Figure 2:  Bland-Altman plots show interobserver agreement for the pharmacokinetic parameters between the two radiologists. Interobserver agreement for K trans
was better than those for ve and vp. SD = standard deviation.

Figure 1:  Definition of the VOI and acquisition of dynamic contrast-enhanced MR imaging–derived pharmacokinetic parameter maps. A, After the VOI was
defined for the enhancing portion, B, the enhancing area was semiautomatically segmented. C, After the identification of the arterial input function (AIF), D, the
concentration-time curve of the tumor was calculated. E, Finally, dynamic contrast-enhanced MR imaging–derived pharmacokinetic parameter maps were obtained.
Gd = gadolinium.

Radiology: Volume 274: Number 3—March 2015  n  radiology.rsna.org 835

NEURORADIOLOGY: Quantitative MR Imaging for Evaluating Progression in Glioblastoma Yun et al

surgical resection followed by con- multivariable stepwise logistic regres- for the use of pharmacokinetic param-
current radiation therapy and chemo- sion analysis showed that mean Ktrans eters in determining the preoperative
therapy with TMZ, we found that the was the only variable that was inde- grade of glioma and for differentiating
pharmacokinetic parameters derived pendently predictive of the differential meningioma from high-grade glioma
from dynamic contrast-enhanced MR diagnosis, although we found a number (20,21), the potential of dynamic con-
imaging data have potential for differ- of significant differences between true trast-enhanced MR imaging–derived
entiating between these groups. Among progression and pseudoprogression pharmacokinetic parameters in dif-
the pharmacokinetic parameters, with univariate analysis. ferentiating between true progression
mean Ktrans and ve in true progression Even though some studies that and pseudoprogression in patients
were significantly higher than those used dynamic contrast-enhanced MR with GBM treated with concurrent
in pseudoprogression. In addition, imaging have shown promising results radiation therapy and chemotherapy
with TMZ has not been previously elu-
cidated, to our knowledge. One group
Figure 3 (22) that used T1-weighted fast spin-
echo dynamic contrast-enhanced MR
imaging to measure the maximum rate
of signal change in 73 patients with ei-
ther recurrent primary glial neoplasms
or radiation necrosis found that recur-
rent gliomas demonstrated a signifi-
cantly elevated maximum rate of signal
change (5.85 sec21 6 1.78) compared
with that in histologically confirmed
radiation necrosis (1.90 sec21 6 0.78).
Another study (23) showed that the
relative peak height and relative per-
centage of signal intensity recovery in
GBM were higher than those in radia-
tion necrosis, implying the potential of
dynamic contrast-enhanced MR imag-
ing–derived pharmacokinetic param-
eters to differentiate recurrent GBM
from radiation necrosis. However, the
findings in the previous studies con-
Figure 3:  Box-and-whiskers graphs show differences in K trans, ve, and vp between the true progression (TP) cerning these pharmacokinetic param-
and pseudoprogression (PP) groups. Mean K trans was higher in the true progression group than in the pseudo- eters might reflect differences in blood
progression group (0.44 min21 6 0.25 [standard deviation] vs 0.23 min21 6 0.10, P = .004). Mean ve was flow rather than permeability.
also higher in the true progression group than in the pseudoprogression group (1.26 6 0.78 vs 0.75 6 0.49, Even though histologic findings of
P = .034). Although mean vp was higher in the true progression group than in the pseudoprogression group, true progression and pseudoprogres-
this difference did not reach significance (0.14 6 0.06 vs 0.11 6 0.06, P = .119). + = Mean, line in box = sion have not been established, at his-
median, boundaries of boxes = interquartile range, whiskers = 95% confidence interval, and  = outlier. topathologic examination, radiation

Table 2
Diagnostic Accuracy of Hemodynamic Variables at Dynamic Contrast-enhanced MR Imaging in the Detection of True Progression
Optimal Threshold Value Positive Predictive Value Negative Predictive
Parameter Median AUC Sensitivity (%) Specificity (%) (%) Value (%)

K trans
 Mean 0.756 (0.575, 0.888) 0.347 59 (33,82) 94 (70, 100) 91 (59, 100) 68 (45, 89)
  10th Percentile 0.801 (0.626, 0.919) 0.070 82 (57, 96) 75 (48, 93) 78 (52, 94) 80 (52, 96)
ve
 Mean 0.750 (0.569, 0.884) 0.570 88 (64, 99) 56 (30, 80) 68 (45, 86) 82 (48, 98)
  5th Percentile 0.813 (0.639, 0.927) 0.182 76 (50, 93) 88 (62, 98) 87 (60, 98) 78 (52, 94)

Note.—Data in parentheses are 95% confidence intervals.

836 radiology.rsna.org  n Radiology: Volume 274: Number 3—March 2015

NEURORADIOLOGY: Quantitative MR Imaging for Evaluating Progression in Glioblastoma Yun et al

necrosis and recurrent GBM are found Figure 4

to be markedly dissimilar. Radiation ne-
crosis is characterized by extensive fi-
brinoid necrosis, vascular dilation, and
endothelial injury to the surrounding
normal cerebral vasculature (24). Re-
current GBM is characterized by vascu-
lar proliferation, manifested by elevated
tumor vasculature density (25–27). In
addition, the presence of a disrupted
neoplastic blood-brain barrier perme-
able to macromolecular contrast agents
is characteristic in recurrent GBM
(25,28,29). In terms of pseudoprogres-
sion, the etiology is thought to be re-
lated to vascular and oligodendroglial
injury leading to inflammation and in-
creased permeability of the blood-brain
barrier (30,31).
The elevated tumor vasculature den-
sity and disrupted blood-brain barrier
contribute to the higher Ktrans and ve in
true progression but not in pseudopro-
gression; Ktrans is the volume transfer
constant between the plasma and ex-
travascular extracellular space, and ve is
the extravascular extracellular space per
unit volume of tissue and is also known
as the leakage space (13,15). According
to the results of our study, Ktrans and ve,
among the dynamic contrast-enhanced
MR imaging–derived pharmacokinetic
parameters, could be used to predict
true progression. Although inflammation
in pseudoprogression may result in in-
creased permeability of the blood-brain
barrier and increased Ktrans and ve, in
our study, in which a relatively short ac-
quisition time (1 minute 30 seconds) for
dynamic contrast-enhanced MR imaging
was used, the results mainly seemed
to reflect the early leakage of contrast
material. According to a previous report Figure 4:  Representative dynamic contrast-enhanced MR imaging–derived pharmacokinetic parameter
(32), recurrent high-grade glioma tends maps and histograms in true progression and pseudoprogression. (a) Images in 79-year-old patient with
to show faster contrast enhancement GBM and true progression. The maps for the pharmacokinetic parameters in the enhancing area show
and leakage than radiation necrosis. increased K trans, ve, and vp values within the enhancing portion compared with the corresponding maps in
Although vp is defined as the blood b. According to findings at follow-up MR imaging after six cycles of adjuvant TMZ (top right MR image), the
patient was confirmed as having true progression. CE = contrast enhanced, T1WI = T1-weighted imaging
plasma volume per unit volume of tissue
(Fig 4 continues).
and can be a marker of angiogenic activ-
ity in a tumor, no significant difference
was observed in vp between true progres- According to histogram analysis of in a left-side shift of the cumulative
sion and pseudoprogression. However, pharmacokinetic parameters, including histograms in the pseudoprogression
the results for interobserver agreement Ktrans and ve, the pseudoprogression group. In addition, the AUCs were high-
of quantitative agreement for the mean group showed a higher frequency of er for low-percentile values of Ktrans and
vp imply that a further validation process voxels with low Ktrans and ve than the ve. This result means that low values of
for measuring vp is necessary. true progression group, which resulted Ktrans and ve within the enhancing areas

Radiology: Volume 274: Number 3—March 2015  n  radiology.rsna.org 837

NEURORADIOLOGY: Quantitative MR Imaging for Evaluating Progression in Glioblastoma Yun et al

Figure 4 (continued) were suggested for true progression.

These studies have revealed sensitivities
(detection of true progression) of 80%–
93.3% and specificities (detection of
pseudoprogression) of 77.8%–100%. In
our study, dynamic contrast-enhanced
MR imaging showed a low sensitivity
of 59% and a high specificity of 94%
when a mean Ktrans of 0.347 min21 was
used as a threshold. We believe that
prospective comparison or a multimo-
dality study is warranted to improve the
diagnostic performance in differentiat-
ing true progression from pseudopro-
gression. While the multiple-flip-angle
method might have taken some time
and been noisy, an estimate could have
been acquired quickly from proton den-
sity–weighted and T1-weighted imaging
where only the repetition time changed
and the echo time was kept constant.
However, the causes of most of the dy-
namic contrast-enhanced imaging er-
rors seem to be related to noise in the
T1 estimates and to physiologic chang-
es in the blood flow from one day to the
next (16). Use of the fixed baseline T1,
as was done in the present study, can
contribute to obtaining more consistent
results (16). However, these character-
istics can result in impossible ve values
greater than 100% (33). Nevertheless,
analysis from the fixed baseline T1
would still be a useful alternative in the
analysis of relative data.
This study had some additional
limitations. First, we used semiauto-
matic segmentation based on contrast-
enhanced T1-weighted images for the
VOI selection to measure the pharma-
cokinetic parameters, a process that
requires dedicated software and fur-
Figure 4 (continued):  (b) Images in 43-year-old patient with GBM and pseudoprogression. Histograms for ther validation. However, a previous
pharmacokinetic parameters in the enhancing area show the left shift of the lines representing relative cu- study (34) revealed that the semiauto-
mulative frequencies for K trans, ve, and vp compared with the corresponding histograms in a. This shift means matic segmentation method based on
that the values in true progression are mainly distributed within the higher values, as compared with those in contrast-enhanced T1-weighted images
pseudoprogression. According to findings at follow-up MR imaging after six cycles of adjuvant TMZ (top right for structural imaging was the most re-
MR image), the patient had pseudoprogression. CE = contrast enhanced, T1WI = T1-weighted imaging. producible for cerebral blood volume
measurement in GBM, as compared
should be focused on during the inter- pseudoprogression by using conven- with other manual segmentation or
pretation of the Ktrans and ve maps for tional MR imaging, diffusion-weighted T2-weighted imaging–based measure-
differentiation between true progres- imaging, and dynamic susceptibility ment methods. Thus, we believe that
sion and pseudoprogression. contrast-enhanced perfusion imaging, our method is also promising for ap-
Recently, several retrospective in which subependymal spread of en- plication in dynamic contrast-enhanced
studies (7–12) have focused on the dif- hancement, low apparent diffusion coef- MR imaging. Second, we did not com-
ferentiation of true progression from ficient, and high cerebral blood volume pare the results with those of other

838 radiology.rsna.org  n Radiology: Volume 274: Number 3—March 2015

NEURORADIOLOGY: Quantitative MR Imaging for Evaluating Progression in Glioblastoma Yun et al
imaging modalities, such as diffusion-
weighted imaging or dynamic suscepti-
bility contrast-enhanced imaging, which
have been known to be promising for
the differentiation between true pro-
gression and pseudoprogression. Thus,
further comparative studies are war-
ranted. Third, the study population
was relatively small; only 17 patients
had true progression, and only 16 had
pseudoprogression.
In conclusion, the present study
demonstrated that dynamic contrast-
enhanced MR imaging–derived phar-
macokinetic parameters, including
Ktrans and ve, in the entire newly devel-
oped or enlarged enhancing lesion may
be useful objective diagnostic tools in
the differentiation of true progression
from pseudoprogression in GBM, as
well as the fact that mean Ktrans is the
most promising parameter for the dif-
ferentiation of true progression from
pseudoprogression.
Disclosures of Conflicts of Interest: T.J.Y. dis-
closed no relevant relationships. C.K.P. disclosed
no relevant relationships. T.M.K. disclosed no
relevant relationships. S.H.L. disclosed no rele-
vant relationships. J.H.K. disclosed no relevant
relationships. C.H.S. disclosed no relevant rela-
tionships. S.H.P. disclosed no relevant relation-
ships. I.H.K. disclosed no relevant relation-
ships. S.H.C. Activities related to the present
article: received a grant from Bayer Healthcare.
Activities not related to the present article: dis-
closed no relevant relationships. Other relation-
ships: disclosed no relevant relationships.
References
1. Stupp R, Mason WP, van den Bent MJ, et
al. Radiotherapy plus concomitant and adju-
vant temozolomide for glioblastoma. N Engl
J Med 2005;352(10):987–996.
2. Erpolat OP, Akmansu M, Goksel F, Bora H,
Yaman E, Büyükberber S. Outcome of newly
diagnosed glioblastoma patients treated by
radiotherapy plus concomitant and adjuvant
temozolomide: a long-term analysis. Tumori
13. Haroon HA, Buckley DL, Patankar TA, et
2009;95(2):191–197.
3. Jeon HJ, Kong DS, Park KB, et al. Clin-
ical outcome of concomitant chemora-
diotherapy followed by adjuvant temo-
zolomide therapy for glioblastaomas:
14. Tofts PS, Brix G, Buckley DL, et al. Esti-
single-center experience. Clin Neurol Neu-
rosurg 2009;111(8):679–682.
4. Brandsma D, Stalpers L, Taal W, Sminia P,
van den Bent MJ. Clinical features, mecha-
nisms, and management of pseudoprogres-
Radiology: Volume 274: Number 3—March 2015  n  radiology.rsna.org 839
sion in malignant gliomas. Lancet Oncol
2008;9(5):453–461.
5. Taal W, Brandsma D, de Bruin HG, et al.
Incidence of early pseudo-progression in a
cohort of malignant glioma patients treated
with chemoirradiation with temozolomide.
Cancer 2008;113(2):405–410.
6. Wen PY, Macdonald DR, Reardon DA, et al.
Updated response assessment criteria for
high-grade gliomas: response assessment in
neuro-oncology working group. J Clin Oncol
2010;28(11):1963–1972.
7. Baek HJ, Kim HS, Kim N, Choi YJ, Kim YJ.
Percent change of perfusion skewness and
kurtosis: a potential imaging biomarker for
early treatment response in patients with
newly diagnosed glioblastomas. Radiology
2012;264(3):834–843.
8. Choi YJ, Kim HS, Jahng GH, Kim SJ, Suh
DC. Pseudoprogression in patients with
glioblastoma: added value of arterial spin
labeling to dynamic susceptibility con-
trast perfusion MR imaging. Acta Radiol
2013;54(4):448–454.
9. Chu HH, Choi SH, Ryoo I, et al. Differen-
tiation of true progression from pseudo-
progression in glioblastoma treated with
radiation therapy and concomitant temo-
zolomide: comparison study of standard and
high-b-value diffusion-weighted imaging. Ra-
diology 2013;269(3):831–840.
10. Lee WJ, Choi SH, Park CK, et al. Diffusion-
weighted MR imaging for the differentiation
of true progression from pseudoprogres-
sion following concomitant radiotherapy
with temozolomide in patients with newly
diagnosed high-grade gliomas. Acad Radiol
2012;19(11):1353–1361.
11. Song YS, Choi SH, Park CK, et al. True
progression versus pseudoprogression in
the treatment of glioblastomas: a compar-
ison study of normalized cerebral blood
volume and apparent diffusion coefficient
by histogram analysis. Korean J Radiol
2013;14(4):662–672.
12. Young RJ, Gupta A, Shah AD, et al. Poten-
tial utility of conventional MRI signs in diag-
nosing pseudoprogression in glioblastoma.
Neurology 2011;76(22):1918–1924.
al. A comparison of Ktrans measurements
obtained with conventional and first pass
pharmacokinetic models in human gliomas.
J Magn Reson Imaging 2004;19(5):527–536.
mating kinetic parameters from dynamic
contrast-enhanced T(1)-weighted MRI of
a diffusable tracer: standardized quan-
tities and symbols. J Magn Reson Imaging
1999;10(3):223–232.
15. Tofts PS, Kermode AG. Measurement of
the blood-brain barrier permeability and
leakage space using dynamic MR imaging.
I. Fundamental concepts. Magn Reson Med
1991;17(2):357–367.
16. Haacke EM, Filleti CL, Gattu R, et al. New
algorithm for quantifying vascular changes
in dynamic contrast-enhanced MRI inde-
pendent of absolute T1 values. Magn Reson
Med 2007;58(3):463–472.
17. Pluim JP, Maintz JB, Viergever MA. Mutual-
information-based registration of medical
images: a survey. IEEE Trans Med Imaging
2003;22(8):986–1004.
18. Sundar H, Shen D, Biros G, Xu C, Da-
vatzikos C. Robust computation of mutual
information using spatially adaptive meshes.
Med Image Comput Comput Assist Interv
2007;10(Pt 1):950–958.
19. Hauck WW, Miike R. A proposal for exam-
ining and reporting stepwise regressions.
Stat Med 1991;10(5):711–715.
20. Roberts HC, Roberts TP, Bollen AW, Ley S,
Brasch RC, Dillon WP. Correlation of mi-
crovascular permeability derived from dy-
namic contrast-enhanced MR imaging with
histologic grade and tumor labeling index: a
study in human brain tumors. Acad Radiol
2001;8(5):384–391.
21. Zhu XP, Li KL, Kamaly-Asl ID, et al. Quanti-
fication of endothelial permeability, leakage
space, and blood volume in brain tumors
using combined T1 and T2* contrast-en-
hanced dynamic MR imaging. J Magn Reson
Imaging 2000;11(6):575–585.
22. Hazle JD, Jackson EF, Schomer DF, Leeds
NE. Dynamic imaging of intracranial lesions
using fast spin-echo imaging: differentiation
of brain tumors and treatment effects. J
Magn Reson Imaging 1997;7(6):1084–1093.
23. Barajas RF Jr, Chang JS, Segal MR, et al.
Differentiation of recurrent glioblastoma
multiforme from radiation necrosis after
external beam radiation therapy with dy-
namic susceptibility-weighted contrast-en-
hanced perfusion MR imaging. Radiology
2009;253(2):486–496.
24. Hopewell JW, Calvo W, Jaenke R, Reinhold
HS, Robbins ME, Whitehouse EM. Micro-
vasculature and radiation damage. Recent
Results Cancer Res 1993;130:1–16.
25. Oh BC, Liu CY, Wang MY, Pagnini PG, Yu C,
Apuzzo ML. Stereotactic radiosurgery: adja-
cent tissue injury and response after high-
dose single fraction radiation. II. Strategies
for therapeutic enhancement, brain injury
mitigation, and brain injury repair. Neu-
rosurgery 2007;60(5):799–814; discussion
799–814.
NEURORADIOLOGY: Quantitative MR Imaging for Evaluating Progression in Glioblastoma Yun et al
26. Sugahara T, Korogi Y, Kochi M, et al. Cor-
relation of MR imaging-determined cerebral
blood volume maps with histologic and an-
giographic determination of vascularity of
gliomas. AJR Am J Roentgenol 1998;171(6):
1479–1486.
27. Wesseling P, Ruiter DJ, Burger PC. Angio-
genesis in brain tumors; pathobiological and
clinical aspects. J Neurooncol 1997;32(3):
253–265.
28. Cha S, Lupo JM, Chen MH, et al. Differen-
tiation of glioblastoma multiforme and single
brain metastasis by peak height and percent-
age of signal intensity recovery derived from
dynamic susceptibility-weighted contrast-en-
hanced perfusion MR imaging. AJNR Am J
Neuroradiol 2007;28(6):1078–1084.
840 radiology.rsna.org  n Radiology: Volume 274: Number 3—March 2015
29. Lupo JM, Cha S, Chang SM, Nelson SJ.
from dynamic contrast-enhanced T1-weight-
Dynamic susceptibility-weighted perfusion ed MR perfusion. Neuro-oncol 2011;13(9):
imaging of high-grade gliomas: character- 1037–1046.
ization of spatial heterogeneity. AJNR Am J
33. Tofts PS, Berkowitz B, Schnall MD. Quan-
Neuroradiol 2005;26(6):1446–1454.
titative analysis of dynamic Gd-DTPA en-
30. Tofilon PJ, Fike JR. The radioresponse
hancement in breast tumors using a per-
of the central nervous system: a dynamic meability model. Magn Reson Med 1995;
process. Radiat Res 2000;153(4):357–370. 33(4):564–568.
34. Jung SC, Choi SH, Yeom JA, et al. Cerebral
31. Wong CS, Van der Kogel AJ. Mechanisms of
radiation injury to the central nervous system: blood volume analysis in glioblastomas using
implications for neuroprotection. Mol Interv dynamic susceptibility contrast-enhanced
2004;4(5):273–284. perfusion MRI: a comparison of manual and
semiautomatic segmentation methods. PLoS
32. Narang J, Jain R, Arbab AS, et al. Differenti- ONE 2013;8(8):e69323.
ating treatment-induced necrosis from recur-
rent/progressive brain tumor using nonmod-
el-based semiquantitative indices derived