Magn Reson Med Sci, Vol. 8, No. 2, pp.

47–53, 2009

MAJOR PAPER

Evaluation of Initial DiŠusion-weighted Image Findings in Acute Stroke

Patients using a Semiquantitative Score
Naomi MORITA1*, Masafumi HARADA2, Masaaki UNO3, Shunji MATSUBARA3,
Shinji NAGAHIRO3, and Hiromu NISHITANI1

Departments of 1Radiology and 2Radiologic Technology, School of Medicine, University of Tokushima

3–18–15, Kuramoto-cho, Tokushima, Tokushima 770–8503, Japan
3Department of Neurosurgery, Institute of Health Biosciences, University of Tokushima Graduate School

(Received June 9, 2008; Accepted January 28, 2009)

Purpose: We evaluated the usefulness of rating diŠusion-weighted images (DWI) using a

semiquantitative score modiˆed from the Alberta Stroke Programme Early CT Score
(ASPECTS) to predict deterioration of neurological symptoms in patients with hyperacute
ischemic stroke who had undergone thrombolytic therapy with recombinant tissue plas-
minogen activator (rt-PA).
Materials and Methods: We examined 22 patients with acute stroke (14 men, 8 women,
mean age 72.5 years) treated with intravenous rt-PA. All were assessed using the National
Institutes of Health Stroke Scale (NIHSS) and underwent emergent magnetic resonance
(MR) imaging within 3 hours and 24 hours of stroke onset. Patients were divided into a de-
teriorated group (16 patients), in which NIHSS scores were increased after thrombolysis,
and a non-deteriorated group (6 patients). We compared the DWI score, ASPECTS, and
volume of hyperintense ischemic lesion on DWI (DWI volume) of the 2 groups and exam-
ined correlations between these scores and initial NIHSS score or DWI volume.
Results: The DWI score and ASPECTS tended to be lower in the deteriorated group
than the non-deteriorated group. In addition, with a cutoŠ valueÃ7, the DWI score could
discriminate the deteriorated group from the non-deteriorated group with a sensitivity of
50z and speciˆcity of 87.5z, whereas for ASPECTS, sensitivity was 50z and speciˆcity,
81.2z. The DWI score, ASPECTS, and DWI volume had no correlation with NIHSS
score but weak negative correlations with the DWI volume ( Pº0.01; Spearman's test).
Comparing initial NIHSS score with each DWI score and DWI volume, the non-deteriorat-
ed group tended to have higher DWI scores and smaller DWI volumes than the deteriorat-
ed group, but there was no statistical diŠerence between initial NIHSS and DWI scores.
Though the DWI score was not statistically diŠerent, the threshold would be set to 6 points
or above. Comparing initial DWI score with volume, patients with low DWI scores tended
to show large variation in DWI volume and patients with small DWI volume showed large
variation in DWI scores. There was no relation between hemorrhagic change and symp-
toms in either group.
Conclusions: The semiquantitative DWI score easily evaluated extent of acute ischemic
lesion on DWI and might be used to predict patient outcome after thrombolytic therapy
more accurately than ASPECTS or DWI volume.

Keywords: acute ischemic stroke, Alberta Stroke Programme Early CT Score, DWI

onset of acute ischemic stroke improves patient

Introduction
outcome.1,2 Treatment response is highly time de-
Thrombolysis with intravenous recombinant tis- pendent; favorable outcome declines rapidly as
sue plasminogen activator (rt-PA) within 3 hours of time from stroke onset increases.2 Early ischemic
changes on noncontrast computed tomography
*Corresponding author, Phone: ＋81-88-633-7173, Fax: ＋81- (CT) have been appreciated and thought to be help-
88-633-7174, E-mail: nmorita＠clin.med.tokushima-u.ac.jp ful for predicting the beneˆt from thrombolysis.3–5

47
48 N. Morita et al.

The extent of ischemic change has been assessed by and hyperintensity less than 1/3 of the MCA terri-
establishing whether hypoattenuation is present in tory on initial DWI. Sites of occlusion on MRA
less or more than one-third of the middle cerebral were: internal carotid artery (ICA), 6; MCA, 13 (8
artery (MCA) territory6 or by using the Alberta M1 occlusion, one M2 occlusion, 4 branch ather-
Stroke Programme Early CT Score (ASPECTS). omatous disease); and unknown, 3 (intact MRA).
ASPECTS is a semiquantitative grading that subdi- The cause of ischemia was atherosclerotic in 11,
vides the MCA territory into 10 regions.7 Extensive cardioembolic in 10, and unknown in one.
ischemic change that aŠect more than one-third of The patients received intravenous administration
the MCA territory or an ASPECTSÃ7 is associ- of 0.6 mg/kg of the rt-PA, alteplase (10z of total
ated with poor outcome and might re‰ect an in- dose as bolus injection during ˆrst minute and
creased risk for thrombolysis-related symptomatic remainder as infusion over 1 hour). The mean time
hemorrhage.6–8 to start rt-PA from symptom onset was 141 min (80
ASPECTS was also applied to assess early ische- to 179 min). Speciˆc guidelines for monitoring and
mic lesion on diŠusion-weighted images (DWI), but treating raised blood pressure for the ˆrst 24 hours
results diŠered from those by ASPECTS on com- after treatment were provided to all patients. Fol-
puted tomography (CT). Recently, the Acute low-up NIHSS and MR imaging examinations were
Stroke Imaging Standardization Group in Japan performed within 24 hours after starting rt-PA
(ASIST-Japan) proposed a semiquantitative score treatment. Existence of intracerebral hemorrhage
modiˆed from the ASPECTS for evaluating early was evaluated by T2*-WI; symptomatic hemor-
ischemic lesions on DWI (DWI score), but its per- rhage was that associated with increase in NIHSS
formance for predicting patient outcome has not more than 4 points.
been fully validated. Therefore, we compared the We divided patients into 2 groups by change be-
use of DWI score with ASPECTS and volume of tween initial and follow-up NIHSS; NIHSS in-
the ischemic lesion on DWI in predicting patient creased in the deteriorated group and decreased or
outcome after acute ischemic stroke and intra- was unchanged in the non-deteriorated group.
venous administration of rt-PA. Two readers with experience using ASPECTS
methodology and who completed a detailed tutorial
evaluated DWI ˆndings using the DWI score and
Materials and Methods
ASPECTS. They were blinded to clinical informa-
From October 1st 2006 to August 31st 2007, we tion, including side of symptoms, and evaluated
analyzed 22 consecutive patients (14 men, 8 wo- scores by consensus. They evaluated the DWI score
men, mean age 72.5 years. [45 to 88 years]) with and ASPECTS using 2 standardized levels, involv-
acute ischemic stroke in the MCA territories who ing the basal ganglionic axial levels and supragan-
were treated with intravenous rt-PA. glionic slices. The DWI score included 11 regions: 3
All patients were assessed using the National In- subcortical structures (lentiform, caudate nucleus,
stitutes of Health Stroke Scale (NIHSS) and under- and posterior rim of internal capsule), 7 cortical
went emergent MR imaging within 3 hours of structures (insula, M1 through M6), and one white
stroke onset. MR imaging was performed on a 3- matter structure (corona radiata); ASPECTS in-
tesla (3T) clinical machine (Signa Echospeed, cluded 10 regions besides the corona radiata (Fig.
General Electronic Medical Systems, Milwaukee 1).
WI, USA) equipped with a standard quadrature In addition, we manually traced the area of
head imaging coil. Our protocol for emergent MR hyperintensity on DWI to measure the DWI vol-
imaging for acute stroke included DWI, MR an- ume in each slice.
giography (MRA), and T2*-WI. Imaging param- We examined the diŠerences in DWI score,
eters of each sequence were: DWI (repetition ASPECTS, DWI volume, NIHSS, and other demo-
time/echo time [TR/TE] 10000/71.8 ms; ˆeld of graphic factors between the deteriorated and non-
vision [FOV] 28; slice thickness/gap 5/1.5 mm; deteriorated groups and compared data between
matrix 128×128; number of excitations [NEX] 1; b the 2 groups using Student's t-test or Wilcoxon
factor 1000 s/mm2 ); MRA (TR/TE 30/3.9 ms, nonparametric test. We also used a cutoŠ value of
FOV 24, slice thickness/gap 1.2/4 mm, ‰ip angle Ã7 to evaluate sensitivities and speciˆcities of DWI
[FA] 159, matrix 384×150); and T2*-WI (TR/TE score and ASPECTS for diŠerentiating the deterio-
380/29 ms, FOV 24, slice thickness/gap 5/1.5 mm, rated from non-deteriorated group because previ-
FA 209 , matrix 512×192). All patients fulˆlled the ous studies associated ASPECTSÃ7 with poor out-
imaging criteria of our institution, including ab- come. Further, we used Spearman's test to analyze
sence of intracranial hemorrhage on initial T2*-WI correlations between the DWI score or ASPECTS

Magnetic Resonance in Medical Sciences

Evaluation of DWI Findings using Semiquantitative Score 49

Fig. 1. Semiquantitative diŠusion-weighted image (DWI) score from modiˆed

Alberta Stroke Programme Early Computed Tomography Score (ASPECTS;
DWI score). The DWI score divided the middle cerebral artery (MCA) territory
into 11 regions, which included white matter. The Acute Stroke Imaging Stand-
ardization Group in Japan ASIST-Japan proposed this scoring system for
evaluating acute ischemia. (http://asist.umin.jp/)

and the DWI volume or NIHSS. DWI volume re‰ected large variation in DWI score
and ASPECTS (Fig. 3).
Results
Discussion
Of 22 patients, six were classiˆed as deteriorated
and 16 as non-deteriorated. Demographic factors ASPECTS is a unique system for evaluating
such as time to start rt-PA and symptomatic/asymp- acute ischemic change and is easily adapted even in
tomatic hemorrhage were not statistically diŠerent emergency situations. A dichotomy of ASPECTS
between the 2 groups. In the deteriorated group, between 0 to 7 and 8 to 10 has been previously vali-
mean age was higher, and NIHSS score and DWI dated on noncontrast CT and shown to have a role
volume were signiˆcantly larger than in the non- in predicting prognosis of patients with acute
deteriorated group ( Pº0.05). The DWI score and stroke who have been treated with intravenous rt-
ASPECTS tended to be lower in the deteriorated PA within 3 hours after stroke onset.7,9 However,
than the non-deteriorated group but were not the ability of ASPECTS to predict patient outcome
statistically diŠerent (Table). With a cutoŠ value of on DWI as well as on CT has not been conˆrmed.
Ã7, the DWI score could discriminate the deterio- Barber and associates reported that ASPECTS
rated group from the non-deteriorated group with values on DWI were lower than those on CT be-
sensitivity of 50z and speciˆcity of 87.5z and cause of the higher sensitivity of DWI than CT in
ASPECTS, with sensitivity of 50z and speciˆcity detecting acute ischemic lesion.10 In this study,
of 81.2z. However, using DWI volume, sensitivity ASPECTS lower than 8 points in 3 of 16 patients in
was 83.3z and speciˆcity, 68.7z with a cutoŠ the non-deteriorated group re‰ected that DWI as
value of Æ20 mL (Fig. 2). compared with CT can overestimate lesion extent.
We found no correlation between DWI score, In contrast, DWI scores less than 8 in 2 patients
ASPECTS, or DWI volume and NIHSS (r＝ suggested that the 11-point system that adds a
－0.28, －0.23, and －0.39; P＝0.20, 0.28, and region of white matter to the original 10 regions in
0.06; Spearman's test). In contrast, we found ASPECTS can compensate for the overestimation
weak, but signiˆcant correlations between the DWI on DWI, although we failed to ˆnd statistical sig-
score or ASPECTS and DWI volume (r＝－0.80 niˆcance between DWI and ASPECT scores. In ad-
and －0.76; Pº0.01 and 0.01; Spearman's test). dition, the overlap in DWI scores between the 2
Patients with low DWI score and ASPECTS patient groups was comparable or smaller when
tended to have largely varying DWI volumes; small compared to the DWI volume, which showed a

Vol. 8 No. 2, 2009

50 N. Morita et al.

Table. Patients treated with recombinant tissue plasminogen activator (rt-PA)

Non-deteriorated group Deteriorated group

P-value
(n＝16) (n＝6)

Mean age (years) 69.8(45–82) 80.0(71–88) 0.028

Gender (male : female) 12 : 04 2 : 04
Occluded artery
Internal carotid artery 1(6.3) 5(83.3)
Middle cerebral artery 12(75.0) 1(16.7)
Unknown 3(18.7) 0(0.0)
Time to rt-PA (minutes) 80–179(142.6) 97–178(137.5) 0.743
Initial NIHSS score 6–20(11.5) 14–23(20.5) 0.010
NIHSS score after treatment 0–19(2.5) 14–40(40.0) º0.001
Alberta Stroke Programme Early CT Score (ASPECTS) 6–10(9.0) 4–10(7.0) 0.055
DiŠusion-weighted imaging (DWI) score 6–11(9.0) 4–11(7.5) 0.087
DWI volume (cc) 0–18.6(5.2) 0–30.6(11.8) 0.030
Recanalization 9(56.3) 1(16.6)
Hemorrhage
symptomatic 1(6.25) 1(16.6)
asymptomatic 7(43.75) 2(33.3)
* Age, time to rt-PA and DWI volume, values in parentheses indicate the average for each group.
* NIHSS, ASPECT score, and DWI score, values in parentheses show the median value.
* Occluded artery, recanalization, and hemorrhage, values in parentheses indicate percentage.

Fig. 2. The diŠusion-weighted imaging (DWI) score, Alberta Stroke Programme Early Computed
Tomography Score (ASPECTS), DWI volume, and initial National Institutes of Health Stroke
Scale (NIHSS) score in patients with acute ischemic stroke. (a) DWI score and NIHSS; (b)
ASPECTS and NIHSS; (c) DWI volume and NIHSS.
With a cutoŠ value of Ã7, the DWI score overestimates 2 patients of the non-deteriorated group
and ASPECTS, 3 patients, though these scores are not signiˆcantly diŠerent between the deteriorat-
ed and non-deteriorated groups. With a cutoŠ value of Æ20 mL, DWI volumes overlapped substan-
tially between the 2 groups, although the diŠerence between the two was signiˆcant. There were no
signiˆcant correlations between the 3 imaging scores and the NIHSS score.

Magnetic Resonance in Medical Sciences

Evaluation of DWI Findings using Semiquantitative Score 51

Fig. 3. Correlation between diŠusion-weighted imaging (DWI) scores or Alberta Stroke Pro-
gramme Early Computed Tomography Score (ASPECTS) and DWI volume. (a) DWI score and
DWI volume; (b) ASPECTS and DWI volume.
There are weak negative correlations between DWI score or ASPECTS and DWI volume ( Pº0.01;
Spearman's test). There was a tendency that patients with small DWI score had large variation of
DWI volume, and patients with small DWI volume showed large variation of DWI score.

Fig. 4. A patient belonging to the non-deteriorated group (67-year-old woman with left M1 occlu-
sion). (a)–(c) Initial magnetic resonance (MR) imaging after 1 hour and 15 min from symptom onset
(d)–(f) Second MR imaging after 7 hours from onset. (a), (b) DiŠusion-weighted imaging (DWI)
showed slightly hyperintense area in left white matter and insula. Initial DWI score was measured to
9 points, and DWI volume was 47.2 cc. (c) Initial MR angiography (MRA) showed left M1 occlu-
sion. After injection, symptoms disappeared rapidly and National Institutes of Health Stroke Scale
(NIHSS) score improved from 20 to 0 points. (d), (e) Hyperintensity area on DWI still remained on
left insula and a small area of high intensity in the white and gray matter, but the lesion did not
spread. (f) Recanalization was conˆrmed by a second MRA. After 2 days, small hypointense lesions
appeared within ischemia on T2*-WI, but these signal changes did not relate to her symptoms (not
shown).

Vol. 8 No. 2, 2009

52 N. Morita et al.

substantial overlap. Hence, DWI score can be uti- DWI score, ASPECTS, or DWI volume and
lized to predict patient outcome presumably more NIHSS, most patients with higher initial NIHSS
accurately than other methods when MR imaging is scores were classiˆed into the deteriorated group
chosen for initial evaluation of patients with acute whether MR imaging ˆndings were good. Higher
stroke. NIHSS would be an important parameter for
Several issues remain with the DWI score. Our predicting prognosis.
results showed no signiˆcant diŠerence between In our case, 11 of 22 patients (50z) showed areas
DWI and original ASPECT scores. Further investi- of hypointensity on T2*-WI after rt-PA therapy,
gation with a larger number of patients is needed to but most were asymptomatic and in only 2 patients
determine an advantage of DWI score over was this related to their symptoms. We felt this sig-
ASPECTS. We should also conˆrm whether the co- nal change showed not only hemorrhage, but also
rona radiata is appropriate as a region for evalua- small leakage of blood product from the vessels as
tion and whether another region is needed to im- a result of their high detectability. However, be-
prove results. Neither was there a consensus as to cause our number of patients was small, we think it
whether small or faint ischemic lesion is counted is di‹cult to evaluate if these scores can predict the
and the non-eloquent regions, such as the insula, hemorrhagic change.
caudate nucleus, and temporal lobe, should be Finally, our study has several limitations. First,
scored equally with the eloquent regions, such as the follow-up period was short. Though we eval-
the corona radiata and internal capsule. Further, uated prognosis at 24 hours after treatment, some
similarly to ASPECTS, the DWI score is utilized cases were recanalized after 24 hours. We guessed
only in the MCA area and cannot be applied to that these patients might have improved long-term
other areas. outcome, and evaluation might be better conducted
Though there was no correlation between the after a longer period. However, because most

Fig. 5. A patient in the deteriorated group (81-year-old woman with left internal carotid artery
[ICA] occlusion). (a)–(c) Initial magnetic resonance (MR) imaging after 1 hour from symptom on-
set. (d)–(f) Second MR imaging after 23 hours from onset. (a), (b) Faint area of hyperintensity on
diŠusion-weighted imaging (DWI) is shown in left temporal lobe (deep to subcortical white matter
and along external capsule). Initial DWI score was measured at 9 points, and DWI volume was 25.3
cc. (c) Initial MR angiography (MRA) showed left ICA occlusion. Though her age was advanced,
intravenous recombinant tissue plasminogen activator (rt-PA) thrombolysis was given after in-
formed consent. (d)–(f) The hyperintense lesion on DWI was spread to the left anterior-temporal
lobe, and the ICA occlusion was unchanged. The patient died after 2 days.

Magnetic Resonance in Medical Sciences

Evaluation of DWI Findings using Semiquantitative Score 53

patients had moved to another hospital after inten- analysis of ATLANTIS, ECASS, and NINDS rt-
sive treatment in the acute phase, follow-up in the PA stroke trials. Lancet 2004; 363:768–774.
chronic phase was di‹cult. Second, our number of 3. Lyden P. Early major ischemic changes on com-
patients was small, and six of 22 had ICA occlu- puted tomography should not preclude use of tis-
sion, a condition generally known to have poor sue plasminogen activator. Stroke 2003; 34:821–
822.
prognosis; this uneven distribution might bias
4. von Kummer R. Early major ischemic changes on
prognosis. By increasing patient number and ex-
computed tomography should preclude use of tis-
tending follow-up time, the DWI score might be sue plasminogen activator. Stroke 2003; 34:820–
demonstrated more useful to evaluate acute is- 821.
chemic lesion. Third, our use of a 3T MR imaging 5. Wardlaw JM, Mielke O. Early signs of brain in-
scanner provided better detection of ischemic lesion farction at CT: observer reliability and outcome
than possible at 1.5T, but higher detectability may after thrombolytic treatment—systematic review.
result in overestimation of the extent of ischemic le- Radiology 2005; 235:444–453.
sion on DWI. Use of a high magnetic ˆeld also 6. von Kummer R, Allen KL, Holle R, et al. Acute
reduces examination time, but we should pay atten- stroke: usefulness of early CT ˆndings before
tion to other diŠerences in using high magnitude thrombolytic therapy. Radiology 1997; 205:327–
MR imaging. 333.
7. Barber PA, Demchuk AM, Zhang J, Buchan AM.
Validity and reliability of a quantitative computed
Conclusion tomography score in predicting outcome of hy-
peracute stroke before thrombolytic therapy.
The semiquantitative DWI score modiˆed from ASPECTS Study Group. Alberta Stroke Program
ASPECTS could distinguish with relatively high Early CT Score. Lancet 2000; 355:1670–1674.
speciˆcity patients with deteriorated NIHSS from 8. Hill MD, Rowley HA, Adler F, et al. Selection of
those without deterioration. This method can be acute ischemic stroke patients for intra-arterial
used to evaluate extent of acute ischemic lesion thrombolysis with pro-urokinase by using
on DWI and may predict more accurately than ASPECTS. Stroke 2003; 34:1925–1931.
ASPECTS or DWI volume patient outcome after 9. Pexman JH, Barber PA, Hill MD, et al. Use of
thrombolytic therapy. the Alberta Stroke Program Early CT Score
(ASPECTS) for assessing CT scans in patients with
acute stroke. AJNR Am J Neuroradiol 2001; 22:
References 1534–1542.
10. Barber PA, Hill MD, Eliasziw M, et al. Imaging of
1. Tissue plasminogen activator for acute ischemic
the brain in acute ischaemic stroke: comparison of
stroke. The National Institute of Neurological Dis-
computed tomography and magnetic resonance
orders and Stroke rt-PA Stroke Study Group. N
diŠusion-weighted imaging. J Neurol Neurosurg
Engl J Med 1995; 333:1581–1587.
Psychiatry 2005; 76:1528–1533.
2. Hacke W, Donnan G, Fieschi C, et al. Association
of outcome with early stroke treatment: pooled

Vol. 8 No. 2, 2009