European Journal of Radiology 74 (2010) e89e94

Contents lists available at ScienceDirect

European Journal of Radiology

journal homepage: www.elsevier.com/locate/ejrad

T2-enhanced tensor diffusion trace-weighted image in the detection of

hyper-acute cerebral infarction: Comparison with isotropic
diffusion-weighted image
Ming-Chung Chou a,b , Wen-Sheng Tzeng c , Hsiao-Wen Chung a,b , Chao-Ying Wang a,b ,
Hua-Shan Liu a,b , Chun-Jung Juan a,b , Chung-Ping Lo d , Chun-Jen Hsueh a , Cheng-Yu Chen a,
a

Department of Radiology, Tri-Service General Hospital, Taipei, Taiwan, ROC

Department of Electrical Engineering, National Taiwan University, Taipei, Taiwan, ROC
c
Department of Radiology, Chi-Mei Medical Center, Tainan, Taiwan, ROC
d
Department of Radiology, Kaohsiung Medical University Chung-Ho Memorial Hospital, Kaohsiung, Taiwan, ROC
b

a r t i c l e

i n f o

Article history:
Received 8 January 2009
Received in revised form 6 April 2009
Accepted 6 April 2009
Keywords:
Tensor diffusion trace-weighted imaging
Acute cerebral infarction

a b s t r a c t
Background and purpose: Although isotropic diffusion-weighted imaging (isoDWI) is very sensitive to the
detection of acute ischemic stroke, it may occasionally show diffusion negative result in hyper-acute
stroke. We hypothesize that high diffusion contrast diffusion trace-weighted image with enhanced T2
may improve stroke lesion conspicuity.
Methods: Five hyper acute stroke patients (M:F = 0:5, average age = 61.8 20.5 y/o) and 16 acute
stroke patients (M:F = 11:5, average age = 67.7 12 y/o) were examined six-direction tensor DWIs at
b = 707 s/mm2 . Three different diffusion-weighted images, including isotropic (isoDWI), diffusion traceweighted image (trDWI) and T2-enhanced diffusion trace-weighted image (T2E trDWI), were generated.
Normalized lesion-to-normal ratio (nLNR) and contrast-to-noise ratio (CNR) of three diffusion images
were calculated from each patient and statistically compared.
Results: The trDWI shows better nLNR than isoDWI on both hyper-acute and acute stroke lesions, whereas
no signicant improvement in CNR. Nevertheless, the T2E trDWI has statistically superior CNR and nLNR
than those of isoDWI and trDWI in both hyper-acute and acute stroke.
Conclusions: We concluded that tensor diffusion trace-weighted image with T2 enhancement is more
sensitive to stroke lesion detection, and can provide higher lesion conspicuity than the conventional
isotropic DWI for early stroke lesion delineation without the need of high-b-value technique.
2009 Elsevier Ireland Ltd. All rights reserved.

1. Introduction
Diffusion-weighted image (DWI) has been proven the most sensitive diagnostic as well as outcome prediction tool for the acute
ischemic stroke in the rst few hours without the need of contrast
agent [110]. Isotropic (traced) DWI is superior to anisotropic DWI
as post-processing isotropic DWIs help to avoid misinterpreting the
stroke lesion by eliminating the diffusion anisotropy from the normal white matter tracts [11]. Notwithstanding, hyper-acute stroke
could exhibit as diffusion negative stroke in isotropic DWI [1216].
Therefore, there is room for improvement in the diffusion imaging method in order to increase the sensitivity of DWI in stroke

This study was supported in part by Chi-Mei National Defense Medical Center
grant CNNDMC9508 and Tri-Service General Hospital grant TSGH-C93-19.
Corresponding author at: Department of Radiology, Tri-Service General Hospital,
325, Cheng-Kung Road, NeiHu, Taipei, Taiwan, ROC.
E-mail address: sandy0928@seed.net.tw (C.-Y. Chen).
0720-048X/$ see front matter 2009 Elsevier Ireland Ltd. All rights reserved.
doi:10.1016/j.ejrad.2009.04.023

lesion depiction. One of this approach is the use of high-b-value DWI

(on the order of b = 20003500 s/mm2 ) to enhance water diffusion
restriction, thus increasing the infarct lesion conspicuity [1719].
However, high-b-value DWI requires high performance gradient
hardware to reach high diffusion sensitivity at short echo time (TE)
and may not be achievable in low gradient MR systems. On the other
hand, it has also been known that acute infarct lesion harbors the
pathological characteristics of cytotoxic edema upon the sodium
pump failure that causes water diffusion restriction and, hence,
shares the long T2 property, though this is usually not obvious in
hyper-acute stage. To achieve high diffusion sensitivity, Tha et al.
took the advantage of long TE (TE = 190 ms) DWI and high-b-value
(3000 s/mm2 ) to further enhance infarct lesions [20]. The combination of high-b-value plus long TE isotropic DWI showed further
improvement on stroke lesion delineation by effectively suppressing the background brain tissues. The expected disadvantage of the
approach is the reduced signal-to-noise ratio (SNR) due to long TE.
The purpose of our study was to increase the diffusion contrast and diffusion sensitivity to cytotoxic edema, and, in the

e90

M.-C. Chou et al. / European Journal of Radiology 74 (2010) e89e94

mean time, suppressing the background brain tissue signal by the

post-processing of diffusion tensor data without the need of high
gradient performance (or high-b-value) on routine MR scanner.

2. Material and methods

2.1. Subjects and data acquisitions

trDWI

Five patients (5 women, mean age = 61.8 20.5 y/o) with

hyper-acute infarction (ictus within 6 h, mean = 4.6 h) and 16
patients (11 men and 5 women, mean age = 67.7 12 y/o) with
acute cerebral infarctions (ictus after 6 h, mean = 46.8 h) were
enrolled in this study. The detailed information on symptom duration and onset time of all stroke patients was listed
in Table 1. None of the patients received acute thrombolytic
therapy. The study was approved by institute review board
and informed consent was obtained from each patient. After
routine T1-weighted (TR/TE/NEX = 600ms/14ms/1), T2-weighted
(TR/TE/ETL/NEX = 4000 ms/99 ms/7/1) and uid attenuated inversion recovery (FLAIR: TR/TI/TE/NEX = 9000 ms/2500 ms/110 ms/1)
images were acquired, six non-collinear DWIs ((1, 1, 0), (1, 1,0), (0,
1, 1), (0, 1, 1), (1, 0, 1), (1, 0, 1)) with b-value = 0 and 707 s/mm2
were subsequently acquired in a 1.5 T MR scanner (Siemens Vision
Plus, Erlangen, Germany) equipped with regular gradient performance (max gradient strength = 25 mT/m). Other parameters for
diffusion imaging were as follows: TR/TE/NEX = 5000 ms/100 ms/4,
matrix size = 128 128, scan time = 2 min and 20 s.
2.2. Post-processing of diffusion-weighted images
All of tensor diffusion images were transferred to a stand-alone
personal computer for image processing using a MATLAB (The
Mathworks, MA, USA) program. Tensor isotropic DWI (isoDWI) was
generated by the following equation [11],
isoDWI

1/2 12

= S0

= S0 exp

DWIx,y DWIx,y DWIy,z DWIy,z DWIx,z DWIx,z

 D +D +D
x
y
z

where, DWIi,j is the diffusion image with diffusion sensitizing gradients applied in (i,j) direction, and S0 is the image acquired without
diffusion sensitizing gradient.
Apart from tensor isotropic DWI, diffusion trace-weighted image
(trDWI) [11] was also calculated according to the following equation,

(1)

1/2 4

= S0

DWIx,y DWIx,y DWIy,z DWIy,z DWIx,z DWIx,z

= S0 exp ((Dx + Dy + Dz ) b)

(2)

Note that in this equation, the exponential term is threefold

as compared with Eq. (1). Since this term is solely responsible for
the diffusion contrast, the DWI with threefold exponential term is
equivalent to DWI acquired at tripled b-value.
Since denominator in Eq. (2) could be a source of noise amplication, in addition to diffusion trace-weighted image, T2-enhanced
diffusion trace-weighted image (T2E trDWI) were also generated to
avoid noise amplication and enhance T2 contrast, in light of the
fact that long T2 cytotoxic edema may occur in ischemic brain tissue
at acute stage, by the following equation,
T 2W trDWI
=


4

DWIx,y DWIx,y DWIy,z DWIy,z DWIx,z DWIx,z

3/2

= S0 exp ((Dx + Dy + Dz ) b)

(3)

In this equation, an order of 3/2 of S0 is used. Since S0 is a T2weighted image, with the order larger than identity, the DWI will
appear higher T2 contrast. In general, the order of S0 can be any
value larger than identity, but it will cause dramatically heavy T2weighted image if larger order of S0 was chosen. For convenience
in this study, we use order of 3/2 for S0 which is directly derived by
taking quarter root of six multiplied DWIs.
2.3. Data analysis
To compare three different DWIs, normalized lesion-to-normal
ratio (nLNR) and contrast-to-noise ratio (CNR) were calculated by

Table 1
Demographics and clinical characteristics of 21 patients with acute ischemic stroke.
Patient no.

Gender

Age (years)

MRA ngings

Timea

Symptom

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21

M
M
M
M
M
F
F
M
F
M
F
M
M
F
F
M
F
M
F
F
F

69
75
61
45
81
26
73
52
47
72
71
80
78
79
73
74
55
77
68
63
74

Negative
R ICA and MCA occlusion
R ICA and MCA occlusion
L M1 Occlusion
R PCA occlusion
L M2 narrowing
R M1 occlusion
L MCA occlusion
L M1/2 and P1/2 narrowing
R MCA and PCA narrowing
L ACA occlusion
L P1 and R M2 narrowing
R MCA occlusion
R ICA and MCA occlusion
Negative
R M2 and P1/2 narrowing
R M2 Occlusion
L ICA and MCA occlusion
R MCA occlusion
L MCA occlusion
L M1 narrowing

42
42
47
39
48
5
4
48
24
30
72
72
66
11
5
72
40
72
24
3
6

Dizziness; poor word comprehension; incoherent speech

L limb weakness; slurred speech; unsteady gait
L limb weakness; slurred speech
R upper limb weakness
L limbs weakness; slurred speech;
R hemiplegia; aphasia
L limb weakness
R limb weakness; slurred speech
L limb weakness, slurred speech
L limb weakness
R limb weakness
L lower limb weakness; alexia
L limb weakness; unstable gait
L limbs weakness; slurred speech
R lower limb weakness; slurred speech
slurred speech; unstable gait
L limb weakness
R hemiplegia
L limb weakness; mental status change
R limb weakness; mental status change
R limb weakness; aphasia

Note: MCA, middle cerebral artery; ICA, internal carotid artery; ACA, anterior cerebral artery; PCA, posterior cerebral artery; M1, M1 segment of MCA; M2, M2 segment of
MCA; P1, P1 segment of posterior cerebral artery; R, right; L, left.
a
Symptom to MRI time (h); MRA, magnetic resonance angiography.

M.-C. Chou et al. / European Journal of Radiology 74 (2010) e89e94

e91

3. Results
3.1. Hyper-acute stroke

Fig. 1. Plot of normalized lesion-to-normal ratio (nLNR) calculated on three different

diffusion-weighted images in ve hyper-acute stroke patients. The P-values were
shown between each comparison pair, and asterisk indicates P < 0.05.

Fig. 2. Plot of contrast-to-noise ratio (CNR) calculated on three different diffusionweighted images in ve hyper-acute stroke patients. The asterisk indicates P < 0.05.

the following two equations,

nLNR =

CNR =

Slesion Snormal
Slesion + Snormal

Slesion Snormal
Noise

(4)

(5)

where, Slesion is the signal intensity of ROI drawn on stroke lesion

(the largest lesion in all slices), Snormal is the signal intensity on
opposite normal tissue with same ROI size, and noise is the standard
deviation of background signal.
Statistical analysis was conducted using paired Students ttest analysis. The paired comparisons include, the CNR and nLNR
between isoDWI and trDWI, the CNR and nLNR between T2E trDWI
and trDWI, and the CNR and nLNR between T2E trDWI and isoDWI.
The difference is considered as signicant if P-value is smaller than
0.05.

The means and standard deviations of nLNR and CNR on three

DWIs and their corresponding P-values were plotted in Figs. 1 and 2,
respectively. In nLNR comparisons, the results showed that trDWI
has higher nLNR than isoDWI with P < 0.05, suggesting the advantage of increasing diffusion exponential term. T2E trDWI also
showed higher nLNR than that isoDWI with P < 0.05. However, no
signicant difference was found between T2E trDWI and trDWI in
hyper-acute stroke patients.
In CNR comparisons, the results showed that CNR in trDWI is
signicantly higher than isoDWI with P < 0.05. Moreover, T2E trDWI
exhibits signicantly higher CNR than both isoDWI and trDWI on
ve hyper-acute stroke patients as shown in Fig. 2.
Fig. 3 shows three different diffusion-weighted images and one
FLAIR image on a representative hyper-acute stroke patient with
right middle cerebral artery occlusion ictus within 3 h. It was
noticed that the ischemic region can only be clearly demarcated
from normal tissue by both trDWI (3b) and T2E trDWI (3c).
Fig. 4 shows three diffusion-weighted images of a 74 y/o stroke
patient with symptom onset within 6 h. In this case, core stroke
lesion was located at slice level (d) and could be visualized in
isoDWI, trDWI and T2E trDWI. However, diffusion-negative stroke
occurs at suburbs of core stroke which cannot be visualized in
isoDWI, but can be well dened in trDWI and T2E trDWI indicated
by white arrows in row (ac) of Fig. 4.
3.2. Acute stroke
Result of statistical comparisons for nLNR and CNR on 16 acute
stroke patients were shown in Figs. 5 and 6, respectively. Similar
to the results of hyper-acute stroke patients, trDWI showed higher
nLNR and CNR than isoDWI (P < 0.05). Different from hyper-acute
stroke lesion, T2E trDWI showed signicantly higher nLNR than that
of trDWI with P < 0.05. Besides, T2E trDWI also showed higher nLNR
and CNR than conventional isoDWI signicantly on acute stroke
patients. Fig. 7 shows three diffusion-weighted images of three
slices from a 77-year-old patient with acute stroke ictus within
72 h. Several small lesions, which showed diffusion negative signal
in isoDWI, were observed in trDWI and T2E trDWI.
4. Discussion
The fundamental concept of hyperacute ischemic stroke at the
molecular level is energy (sodium pump) failure, leading to an
increase of water inux into the cell that further causes cell edema
and restriction of water diffusion in the insinuated extra-cellular
spaces. Therefore, any MR imaging sequence that can enhance
the detection of the very early cell edema and water diffusion
restriction will be the one suitable for acute stroke imaging. In this

Fig. 3. Three different diffusion-weighted images and one FLAIR image of a 63 years old female stroke patient with symptom onset within 3 h. The images from the left to
right are trace ADC (a), conventional isoDWI (b), trDWI (c), T2E trDWI (d) and FLAIR (e). White arrows point out the margin of stroke lesion which cannot be delineated by
isoDWI (b), but clearly dened by both trDWI (c) and T2E trDWI (d).

e92

M.-C. Chou et al. / European Journal of Radiology 74 (2010) e89e94

Fig. 4. Four slices of trace ADC map and three different diffusion-weighted images of a 74 years old stroke patient with symptom onset within 6 h. In this case, the core
stroke lesion, which is located in slice (d), can be visualized by trace ADC map (1st column), isoDWI (2nd column), trDWI (3rd column) and T2E trDWI (4th column). However,
the diffusion negative signal, which occurs at the suburbs of central stroke lesion in isoDWIs (rows (a)(c) of 2nd column), could be visualized in trDWIs (rows (a)(c) of 3rd
column) and T2E trDWIs (rows (a)(c) of 4th column) indicated by white arrows.

study, we employed six-direction tensor diffusion trace-weighted

images with or without T2 enhancement to compare with conventional isoDWI in hyper-acute and acute stroke patient groups.
We found that T2-enhanced diffusion trace-weighted image shows
signicantly higher nLNR and CNR on both hyper-acute and acute

stroke patient groups. However, to achieve signicant, not quantum, imaging improvement for DWI, a very high-b-value up to
20003500 s/mm2 is usually required [1719]. This is usually at
the expense of signal to noise ratio and is not always applicable
in routine MR scanner due to the limited gradient strength.

Fig. 5. Plot of normalized lesion-to-normal ratio (nLNR) values calculated on three

different diffusion-weighted images in 16 acute stroke patients. The asterisk indicates P < 0.05.

Fig. 6. Plot of contrast-to-noise ratio (CNR) calculated on three different diffusionweighted images in 16 acute stroke patients. The asterisk indicates P < 0.05.

M.-C. Chou et al. / European Journal of Radiology 74 (2010) e89e94

e93

Fig. 7. Trace ADC maps and three diffusion-weighted images of three slices from a 77-year-old stroke patient with symptom onset within 72 h. In this case, the scattered
stroke lesions of the rst slice were visualized in trace ADC map (1st column), isoDWI (2nd column), trDWI (3rd column) and T2E trDWI (4th column). However, several small
lesions, showing diffusion negative signal in isoDWI (rows (b) and (c)), were observed in trDWI and T2E trDWI as indicated by white arrows.

This study utilizes tripled diffusion exponential term in diffusion

trace-weighted image processing to enhance the diffusivity sensitivity with diffusion contrast equivalent to the DWI acquired with
a threefold b-value, or b = 707 3 s/mm2 . And our study demonstrated that trDWI exhibits higher nLNR (or diffusion contrast) than
that of conventional isoDWI on both hyper-acute and acute stroke
patients if noise issue was ignored. As noise was taken into account,
trDWI however shows similar CNR to that of isoDWI. This is because
that noise level was also elevated during trDWI calculation.
By eliminating noise amplication, T2E trDWI shows substantially higher CNR on both hyper-acute and acute stroke patient
groups than trDWI, but it exhibits relatively higher T2 contrast
than trDWI. With increased T2 and diffusion contrast, T2E trDWI
suppresses background tissue more than isoDWI, and therefore
conrms the concept in a prior study that using long TE and high-bvalue techniques in diffusion-weighted image is helpful to suppress
background tissue [21].
Besides, T2E trDWI helps contrast stroke lesion (nLNR) about
10% at acute stage as compared with trDWI, whereas no signicant
increase was observed at hyper-acute stage. This nLNR improvement on acute stroke patients could be attributed to the fact that
cytotoxic edema has already begun after rst 6 h [20] which, however, is hardly observed in FLAIR images.
Contrast-to-noise ratio (CNR) was widely used to evaluate the
sensitivity of diffusion-weighted images in detecting ischemic
lesions [2224]. This study also used nLNR to compare the diffusion contrast of three diffusion-weighted images such that we
can emphasize the advantage of using tripled diffusion exponential
term.
Since the diffusion data in this study were acquired at b-value
of 750 s/mm2 , the isoDWI exhibited somewhat different diffusion

contrast from the trDWI and T2E trDWI, especially in cerebral ventricles showing the intermediate signal intensity, which caused the
adjacent tiny stroke lesions inconspicuous in isoDWI, as shown in
Fig. 7(rows (b) and (c)). By taking advantage of increasing diffusion
exponential term, the trDWI and T2E trDWI were both capable of
contrasting tiny stroke lesions in the vicinity of cerebral ventricles.
However, it should be noticed that although our study showed
that T2E trDWI helps improve CNR by utilizing heavily T2 and diffusion weightings in diagnosis of cerebral infarct, the trace ADC
maps, i.e. sum of three orthogonal diffusivities, of isoDWI, trDWI and
T2E trDWI were numerically identical. Therefore, trace ADC maps
of isoDWI, trDWI and T2E trDWI were not compared in this study.
5. Conclusions
In this study, we demonstrated that tensor trDWI helps suppress background tissue and therefore enhances stroke lesion on
both acute and hyper-acute stroke patients. With T2 enhancement,
T2E trDWI helps improve the conspicuity of acute stroke lesions
without the need of high gradient strength (or high-b-value) in
routine MR scanners.
References
[1] Gonzalez RG, Schaefer PW, Buonanno FS, et al. Diffusion-weighted MR imaging:
diagnostic accuracy in patients imaged within 6 h of stroke symptom onset.
Radiology 1999;210:15562.
[2] Singer MB, Chong J, Lu DF, Schonewille WJ, Tuhrim S, Atlas SW. Diffusionweighted MRI in acute subcortical infarction. Stroke 1998;29:1336.
[3] Warach S, Chien D, Li W, Ronthal M, Edelman RR. Fast magneticresonance diffusion-weighted imaging of acute human stroke. Neurology
1992;42:171723.

e94

M.-C. Chou et al. / European Journal of Radiology 74 (2010) e89e94

[4] Hand PJ, Wardlaw JM, Rivers CS, et al. MR diffusion-weighted imaging and
outcome prediction after ischemic stroke. Neurology 2006;66:115963.
[5] Kuhl CK, Textor J, Gieseke I, et al. Acute and subacute ischemic stroke at
high-eld-strength (30-T) diffusion-weighted MR imaging: intraindividual
comparative study. Radiology 2005;234:50916.
[6] Singhal AB, Topcuoglu MA, Koroshetz WJ. Diffusion MRI in three types of anoxic
encephalopathy. Journal of the Neurological Sciences 2002;196:3740.
[7] Arbelaez A, Castillo M, Mukherji SK. Diffusion-weighted MR imaging of global
cerebral anoxia. American Journal of Neuroradiology 1999;20:9991007.
[8] Pereira RS, Harris AD, Sevick RJ, Frayne R. Effect of b value on contrast during
diffusion-weighted magnetic resonance imaging assessment of acute ischemic
stroke. Journal of Magnetic Resonance Imaging 2002;15:5916.
[9] van Everdingen KJ, van der Grond J, Kappelle LJ, Ramos LMP. Mali WPTM.
Diffusion-weighted magnetic resonance imaging in acute stroke. Stroke
1998;29:178390.
[10] Warach S, Gaa J, Siewert B, Wielopolski P, Edelman RR. Acute human stroke
studied by whole-brain echo-planar diffusion-weighted magnetic-resonanceimaging. Annals of Neurology 1995;37:23141.
[11] Chong JN, Lu DF, Aragao F, et al. Diffusion-weighted MR of acute cerebral
infarction: comparison of data processing methods. American Journal of Neuroradiology 1998;19:17339.
[12] Oppenheim C, Stanescu R, Dormont D, et al. False-negative diffusion-weighted
MR ndings in acute ischemic stroke. American Journal of Neuroradiology
2000;21:143440.
[13] Ay H, Buonanno FS, Sorenson G, et al. Negative diffusion-weighted MR imaging
(DWI) with stroke-like focal decits. Stroke 1998;29:3071307.
[14] Crozier S, Oppenheim C, Stanescu R, et al. Early negative diffusion
MRI (DWI) in ischemic vertebro-basilar stroke. Neurology 2000;54:A94
194.

[15] Pedraza S, Osuna T, Davalos A, Teruel J, Vera-Sancho J, Inaraja L. False negative

diffusion in acute ischemic stroke. Revista De Neurologia 2002;34:11279.
[16] Wei W, Goldstein S, Scheuer ML, Branstetter BF. Acute stroke syndrome with
xed neurological decit and false-negative diffusion-weighted imaging. Journal of Neuroimaging 2003;13:15861.
[17] Meyer JR, Gutierrez A, Mock B, et al. High-b-value diffusion-weighted MR
imaging of suspected brain infarction. American Journal of Neuroradiology
2000;21:18219.
[18] DeLano MC, Cooper TG, Siebert JE, Potchen MJ, Kuppusamy K. High-b-value
diffusion-weighted MR imaging of adult brain: image contrast and apparent diffusion coefcient map features. American Journal of Neuroradiology
2000;21:18306.
[19] Kim HJ, Choi CG, Lee DH, Lee JH, Kim SJ, Suh DC. High-b-value diffusionweighted MR imaging of hyperacute ischemic stroke at 1.5 T. American Journal
of Neuroradiology 2005;26:20815.
[20] Tha KK, Terae S, Yamamoto T, et al. Early detection of global cerebral anoxia:
improved accuracy by high-b-value diffusion-weighted imaging with long echo
time. American Journal of Neuroradiology 2005;26:148797.
[21] Loubinoux I, Volk A, Borredon J, et al. Spreading of vasogenic edema and cytotoxic edema assessed by quantitative diffusion and T2 magnetic resonance
imaging. Stroke 1997;28:41926.
[22] Kosior R, Wright C, Kosior J, et al. 3-Tesla versus 1.5-Tesla magnetic resonance
diffusion and perfusion imaging in hyperacute ischemic stroke. Cerebrovascular
Diseases 2007;24:3618.
[23] Pereira R, Harris A, Sevick R, et al. Effect of b value on contrast during diffusionweighted magnetic resonance imaging assessment of acute ischemic stroke.
Journal of Magnetic Resonance Imaging 2002;5:5916.
[24] Yamada K, Kubota H, Kizu O, et al. Effect of intravenous gadoliniumDTPA on
diffusion-weighted images. Stroke 2002;33:1799802.