7
ORIGINAL ARTICLE Prediction of Poor Outcome in Cerebellar Infarction by Diffusion MRI Zahari Tchopev Marc Hiller Jiachen Zhuo Joshua Betz Rao Gullapalli Kevin N. Sheth Published online: 7 August 2013 Ó Springer Science+Business Media New York 2013 Abstract Objective Identification of patients with posterior fossa infarction at risk for neurological deterioration remains a challenge. MRI-based assessments of MCA infarction can predict poor outcome. Similar quantitative imaging mea- sures after cerebellar stroke have not been studied. We tested the hypothesis that MRI-based volumetric assess- ment of cerebellar infarcts can provide reliable information for the prediction of poor outcome. Design We retrospectively identified 44 consecutive subjects (age 55.2 ± 13) with cerebellar stroke who underwent MRI with diffusion-weighted imaging (DWI) (median 63.7 h). Subjects were divided into poor (n = 13) and good outcomes (n = 31). Poor outcome was defined as having at least one of the following criteria: (1) mortality, (2) decompressive craniectomy, (3) ventriculostomy, and (4) decrease level of consciousness. DWI and cerebellar volume were defined on apparent diffusion coefficient maps. The ratio of the lesion volume to the whole cere- bellum volume was calculated (rVolume). Measurements and Main Results Logistic regression revealed that lesion volume and rVolume were associated with increased risk of poor outcome, even after adjusting for age and NIHSS (v 2 = 8.2230, p < 0.0042; v 2 = 8.3992, p < 0.0038, respectively). The receiver operating charac- teristic curve with age, NIHSS, and volume or rVolume achieved an AUC of 0.816 (95 % CI 0.678–0.955) and 0.831 (95 % CI 0.6989–0.9636), respectively. Conclusions Quantitative volumetric measurement pre- dicts poor outcome of cerebellar stroke patients, even when controlling for age and NIHSS. Quantitative analysis of diffusion MRI may assist in identification of patients with cerebellar stroke at highest risk of neurological deteriora- tion. Prospective validation is warranted. Keywords Stroke Á Cerebellum Á Magnetic resonance imaging Á Outcome assessment Á Stroke volume Introduction Life-threatening edema causing brainstem compression and obstructive hydrocephalus following cerebellar stroke occurs in about 10–20 % of patients [14]. The mortality rate of patients who develop post-ischemic cerebellar edema is an estimated 40 % while 50 % of survivors have disabling deficits even after surgical intervention [5]. Prompt identification of patients with cerebellar stroke is a challenge [69]. A parallel consequence is the difficulty in identifying cerebellar stroke patients at high risk for Z. Tchopev University of Maryland College Park, College Park, MD, USA M. Hiller University of Maryland School of Medicine, Baltimore, MD, USA J. Zhuo Á R. Gullapalli Magnetic Resonance Research Center, Department of Diagnostic Radiology and Nuclear Medicine, University of Maryland School of Medicine, Baltimore, MD, USA J. Betz Department of Biostatistics, Johns Hopkins Bloomberg School of Public Health, Baltimore, MD, USA K. N. Sheth (&) Division of Neuro-Critical Care & Emergency Neurology, Yale University School of Medicine, 628 S. Wolfe Street, Baltimore, MD 21231, USA e-mail: [email protected] 123 Neurocrit Care (2013) 19:276–282 DOI 10.1007/s12028-013-9886-2

Prediction of Poor Outcome in Cerebellar Infarction by Diffusion MRI

  • Upload
    kevin-n

  • View
    213

  • Download
    1

Embed Size (px)

Citation preview

Page 1: Prediction of Poor Outcome in Cerebellar Infarction by Diffusion MRI

ORIGINAL ARTICLE

Prediction of Poor Outcome in Cerebellar Infarctionby Diffusion MRI

Zahari Tchopev • Marc Hiller • Jiachen Zhuo •

Joshua Betz • Rao Gullapalli • Kevin N. Sheth

Published online: 7 August 2013

� Springer Science+Business Media New York 2013

Abstract

Objective Identification of patients with posterior fossa

infarction at risk for neurological deterioration remains a

challenge. MRI-based assessments of MCA infarction can

predict poor outcome. Similar quantitative imaging mea-

sures after cerebellar stroke have not been studied. We

tested the hypothesis that MRI-based volumetric assess-

ment of cerebellar infarcts can provide reliable information

for the prediction of poor outcome.

Design We retrospectively identified 44 consecutive

subjects (age 55.2 ± 13) with cerebellar stroke who

underwent MRI with diffusion-weighted imaging (DWI)

(median 63.7 h). Subjects were divided into poor (n = 13)

and good outcomes (n = 31). Poor outcome was defined as

having at least one of the following criteria: (1) mortality,

(2) decompressive craniectomy, (3) ventriculostomy, and

(4) decrease level of consciousness. DWI and cerebellar

volume were defined on apparent diffusion coefficient

maps. The ratio of the lesion volume to the whole cere-

bellum volume was calculated (rVolume).

Measurements and Main Results Logistic regression

revealed that lesion volume and rVolume were associated

with increased risk of poor outcome, even after adjusting for

age and NIHSS (v2 = 8.2230, p < 0.0042; v2 = 8.3992,

p < 0.0038, respectively). The receiver operating charac-

teristic curve with age, NIHSS, and volume or rVolume

achieved an AUC of 0.816 (95 % CI 0.678–0.955) and 0.831

(95 % CI 0.6989–0.9636), respectively.

Conclusions Quantitative volumetric measurement pre-

dicts poor outcome of cerebellar stroke patients, even when

controlling for age and NIHSS. Quantitative analysis of

diffusion MRI may assist in identification of patients with

cerebellar stroke at highest risk of neurological deteriora-

tion. Prospective validation is warranted.

Keywords Stroke � Cerebellum �Magnetic resonance imaging � Outcome assessment �Stroke volume

Introduction

Life-threatening edema causing brainstem compression and

obstructive hydrocephalus following cerebellar stroke

occurs in about 10–20 % of patients [1–4]. The mortality

rate of patients who develop post-ischemic cerebellar

edema is an estimated 40 % while 50 % of survivors have

disabling deficits even after surgical intervention [5].

Prompt identification of patients with cerebellar stroke is a

challenge [6–9]. A parallel consequence is the difficulty in

identifying cerebellar stroke patients at high risk for

Z. Tchopev

University of Maryland College Park, College Park, MD, USA

M. Hiller

University of Maryland School of Medicine, Baltimore,

MD, USA

J. Zhuo � R. Gullapalli

Magnetic Resonance Research Center, Department of Diagnostic

Radiology and Nuclear Medicine, University of Maryland

School of Medicine, Baltimore, MD, USA

J. Betz

Department of Biostatistics, Johns Hopkins Bloomberg School

of Public Health, Baltimore, MD, USA

K. N. Sheth (&)

Division of Neuro-Critical Care & Emergency Neurology,

Yale University School of Medicine, 628 S. Wolfe Street,

Baltimore, MD 21231, USA

e-mail: [email protected]

123

Neurocrit Care (2013) 19:276–282

DOI 10.1007/s12028-013-9886-2

Page 2: Prediction of Poor Outcome in Cerebellar Infarction by Diffusion MRI

neurological deterioration, which may impede appropriate

initiation of medical interventions that prevent irreversible

secondary damage or death [5, 10]. Prognostic factors for

patient treatment selection remain unclear in severe cere-

bellar stroke [11]. In patients with large cerebellar stroke

with swelling, early surgical management prior to poten-

tially fatal complications may lead to a better outcome

[1, 3, 4, 10, 12, 13].

The most commonly used brain imaging test for stroke

is computed tomography (CT), which is widely available

but often not sensitive for the detection of ischemia in the

first hours after symptom onset [14]. Initial CT imaging is

normal in 25 % of patients with cerebellar stroke who go

on to develop mass effect [2]. Magnetic resonance imaging

(MRI) is far more sensitive than CT in diagnosis of acute

ischemic stroke, particularly in the posterior fossa [15, 16].

Recent retrospective and prospective studies have

established the utility of MRI to predict a poor outcome in

middle cerebral artery infarction using baseline infarct

volume measurements with high specificity [17–20]. End-

points in these studies included neurological deterioration,

such as level of arousal, radiological findings, surgery,

death, or any combination. Cohort sizes in the retrospective

studies were 28, 37, and 38 [17, 19, 20]. These studies

suggested that quantitative MRI may help in guiding early

treatment decisions and monitoring. To our knowledge, no

study has systematically evaluated MRI-based volumetric

findings and their ability to predict neurological deterio-

ration in the setting of cerebellar infarction [2, 21]. Our aim

was to evaluate and identify early imaging characteristics

that predict poor outcome secondary to neurological dete-

rioration after cerebellar infarction. We tested the

hypothesis that quantitative MRI parameters may predict

neurological deterioration in patients with posterior fossa

infarction.

Subjects and Methods

Subjects

This retrospective study consisted of 44 consecutive sub-

jects (30 male and 14 female) with a posterior circulation

infarct who underwent diffusion-weighted imaging (DWI)

MRI. Subjects presented to a tertiary academic stroke

center between January 2007 and July 2011. Subjects with

cerebellar infarction were identified from the hospital

stroke registry. A vascular neurologist, KS, reviewed all

MRIs and selected those with restricted diffusion in the

cerebellum. We included all subjects who had (1) an

infarction of the cerebellum, (2) a time of symptom onset,

(3) neurological examination, (4) and a DWI performed

after symptom onset. Additional demographic and clinical

variables were collected. Severity of neurological deficit at

admission was determined from the National Institutes of

Health Stroke Scale score (NIHSS).

All subjects were admitted to the primary stroke center

and were treated according to standardized protocols.

Subjects were monitored for clinical deterioration in the

stroke or neuroscience intensive care unit. Neurosurgery

was consulted for evaluation at the discretion of the inpa-

tient stroke attending, and the decision to proceed with

surgery was made jointly, including family representatives,

in all cases. In the event of change in level of arousal,

follow-up neuroimaging was performed. Patients were kept

eunatremic. Osmotherapy and intubation were used in the

setting of neurological deterioration.

End Point

Subjects were classified as poor outcome if they met at least

one of the following criteria: (1) in-patient mortality, (2)

decompressive craniectomy, (3) ventriculostomy, and (4)

decrease of consciousness to a score of 1 or greater on item

1a of NIHSS where no other cause of secondary deterio-

ration was noted [1, 4, 11, 12, 22–25]. The criteria were

chosen as potential surrogates for neurological deteriora-

tion. Decompressive craniectomy and ventriculostomy, as

well as all deaths, were made in the context of neurological

deterioration. Determination of neurological deterioration

during hospital course was based on NIHSS as documented

in the medical chart. Item 1a of the NIHSS evaluates a

patient’s level of consciousness. An increase in the NIHSS

has been associated with an increased risk of neurological

and medical complications in posterior circulation strokes

and an eligibility criteria for decompressive surgery in

patients with malignant middle cerebral artery infarction

[26, 27]. Outcome assessment was performed independent

of and blinded from analysis of MRI examinations.

Diffusion-Weighted Magnetic Resonance Imaging

MRI examinations were performed at a median of 41.1 h

after the onset of acute neurological symptoms. All MRI

examinations were performed using either a Siemen Av-

anto 1.5T (n = 40) or a Siemens Tim Trio 3T (n = 4) MRI

system (Siemen Medical Solutions, PA). The diffusion-

weighted (DW) images obtained on the 1.5T consisted of

26 slices covering the entire brain at a slice thickness of

5 mm with a 1-mm gap. The field of view (FoV) used was

either 220 or 230 mm at an acquisition matrix of

134 9 192 pixels. The 3T differed with a FoV of 218 by

240 at a matrix of 174 9 192 pixels consisting of 28 slices.

All DW images were obtained in the axial plane using dual

spin echo planar imaging (EPI) method with diffusion

Neurocrit Care (2013) 19:276–282 277

123

Page 3: Prediction of Poor Outcome in Cerebellar Infarction by Diffusion MRI

weighting in three orthogonal directions at a b-value of

1000 s mm2. The echo-time (TE) for the DW images was

86 ms with a repetition time (TR) of 5400 and 89 ms with

a TR of 4500 for the 1.5T and 3T scans, respectively.

Volume Measurement and VOI Analysis

Volumes and apparent diffusion coefficient (ADC) values

were measured from ADC maps prior to determination of

subject outcome. First, regions immediately adjacent to, or

when necessary lateral to, lesions on each slice were

manually outlined to obtain the mean ADC value of normal

cerebellar tissue. Voxels with ADC values above

1 9 10-3 mm2 s were excluded to reduce signal contam-

ination from cerebrospinal fluid. The average voxel

intensity for the volume of interest ‘‘VOI’’ was recorded. A

similar procedure was used to obtain volumetric and mean

ADC measurements of the whole cerebellum. Lesions were

first manually outlined on the ADC map to obtain a VOI. A

further threshold was used to include only voxels with

ADC values that are most likely to represent affected tis-

sue. This threshold was set at the normal tissue average

ADC minus one standard deviation. The whole cerebellum

was manually traced and a threshold was set at

1 9 10-3 mm2 s. The whole cerebellar average ADC and

the standard deviation and volume were recorded. All

volume measurements were determined by integration

across all relevant slices. Figure 1 is a visual representation

of the outlining procedure. The ratio of the lesion volume

to the whole cerebellum volume was calculated for all the

lesion thresholds (rVolume) as well as the ratio of the

lesion mean ADC to normal cerebellum mean ADC

(rADC).

Statistical Analysis

Univariate associations between covariates and outcomes

were evaluated using the Wilcoxon Rank Sum Test or

Fisher’s Exact Test, where appropriate. To test the primary

hypothesis, we performed a multivariable logistic regres-

sion analysis with poor outcome as the dependent variable.

Imaging markers were entered into stepwise selection

logistic regression model, adjusted for age and NIHSS.

For logistic regression models, the area under the

receiver operating characteristic (ROC) curve was calcu-

lated using the method of DeLong [28]. Stepwise logistic

regression was performed using SAS 9.2 (SAS Institute

Inc., Cary, NC). ROC analyses were conducted in R 2.15

(R Foundation for Statistical Computing, Vienna, Austria.)

using pROC 1.5.1 [29]. We further calculated sensitivity,

specificity, negative predictive value (NPV), positive pre-

dictive value (PPV), true positives (TP), and false positives

(FP) at a range of thresholds for volume of infarct and

rVolume. The fitted probability of poor outcome from the

logistic regression model was calculated at a range of

probabilities that included age, NIHSS, and rVolume.

Statistical significance was judged at a threshold of

p < 0.05.

Results

Within the study period, 44 subjects with acute ischemic

stroke met the inclusion criteria. Demographic and clinical

features are summarized in Table 1. 13 (29.5 %) subjects

met poor outcome criteria. The NIHSS (p = 0.0845) and

time to scan (p = 0.0597) approached significance. There

was a significant difference in subjects with a history of

coronary artery disease (CAD)/prior myocardial infarction

(p < 0.0001) and the cerebellar vascular territory affected

(p = 0.0131). Quantitative MRI variables are summarized

in Table 2. Poor outcomes were associated with higher

infarct volume (p = 0.003), lower whole cerebellar aver-

age ADC (p = 0.0279), higher whole cerebellar average

ADC standard deviation (p = 0.0162), and larger rVolume

(p = 0.009).

The rVolume was selected in stepwise logistic regres-

sion as the most important prognostic imaging

measurement and consequently used in a logistic regres-

sion model containing NIHSS and age. Lesion volume was

simultaneously tested due to high correlation with rVolume

(r = 0.96). Logistic regression revealed that lesion volume

Fig. 1 Outlining of cerebellum. The whole cerebellum (blue), infarct

(yellow) and non-infarct tissue (red) were outlined on DWI-derived

ADC maps (Color figure online)

278 Neurocrit Care (2013) 19:276–282

123

Page 4: Prediction of Poor Outcome in Cerebellar Infarction by Diffusion MRI

and rVolume were associated with increased risk of poor

outcome, even after adjusting for age and NIHSS

(v2 = 8.2230, p < 0.0042 and v2 = 8.3992, p < 0.0038;

respectively). The logistic regression model adjusted for

age and NIHSS achieved an AUC of 0.816 (95 % CI

0.678–0.955) for lesion volume and 0.831 (95 % CI

0.6989–0.9636) for rVolume, respectively. The ROC

curves for the logistic regression models of poor outcome

are illustrated in Fig. 2a. rVolume alone performed better

than NIHSS alone; however, the model that included

rVolume and NIHSS performed the best with no

improvement when age was included. Figure 2b lists the

characteristics of the ROC analysis for rVolume at a range

of sensitivities.

Table 1 Demographic and clinical features

Cohort (n = 44) Good outcome (n = 31) Poor outcome (n = 13) p value

Age (years) 55.2 ± 13 53.5 ± 13.6 59.2 ± 10.8 0.20

Male gender 30 (68.2 %) 19 (61.3 %) 11 (84.6 %) 0.17

Race (caucasian) 20 (45.5 %) 16 (51.6 %) 4 (30.8 %) 0.32

NIH stroke scale score 5.2 ± 4.7 4.3 ± 3.5 7.4 ± 6.4 0.08

Time to scan (h) 63.7 ± 65.3 74.9 ± 71.5 37.1 ± 37.5 0.06

Atrial Fib. 3 (6.8 %) 1 (3.2 %) 2 (15.4 %) 0.15

Hypertension 25 (56.8 %) 18 (58.1 %) 7 (53.8 %) 1

Diabetes 15 (34.1 %) 12 (38.7 %) 3 (23.1 %) 0.49

Brainstem infarction 13 (29.5 %) 8 (25.8 %) 5 (38.5 %) 0.40

CAD/Prior MI 6 (13.6 %) 0 (0 %) 6 (46.2 %) <0.0001

PVD 1 (2.3 %) 0 (0 %) 1 (7.7 %) 0.25

Hyperlipidemia 10 (22.7 %) 7 (22.6 %) 3 (23.1 %) 0.69

Previous TIA 3 (6.8 %) 1 (3.2 %) 2 (15.4 %) 0.15

Previous stroke 9 (20.5 %) 6 (19.4 %) 3 (23.1 %) 0.67

Smoker 20 (45.5 %) 15 (48.4 %) 5 (38.5 %) 1

CHF 1 (2.3 %) 0 (0 %) 1 (7.7 %) 0.25

Nystagmus mention 5 (11.4 %) 3 (9.7 %) 2 (15.4 %) 1

Occlusion 0.57

Basilar 2 (4.5 %) 0 (0 %) 2 (15.4 %)

Vertebral 7 (15.9 %) 5 (16.1 %) 2 (15.4 %)

Vertebrobasilar 6 (13.6 %) 4 (12.9 %) 2 (15.4 %)

None 29 (65.9 %) 22 (71.0 %) 7 (53.8 %)

Cerebellar vascular territorya 0.01

AICA 2 (4.5 %) 2 (6.5 %) 0 (0 %)

PICA 23 (52.3 %) 18 (58.1 %) 5 (38.5 %)

SCA 10 (32.3 %) 8 (25.8 %) 2 (15.4 %)

Multiple 9 (20.5 %) 3 (9.7 %) 6 (46.2 %)

a AICA anterior inferior cerebellar artery, PICA posterior inferior cerebellar artery, SCA superior cerebellar artery

Table 2 Quantitative MRI measurements

Variable Cohort (n = 44) Good outcome (n = 31) Poor outcome (n = 13) p value

Infarct volumea 16.6 ± 17.4 11.7 ± 13.5 28.4 ± 20.5 0.003

Infarct ADCb 501.0 ± 67.6 508.6 ± 59.8 482.9 ± 83.1 0.252

Cerebellar volumea 109.6 ± 25.7 105.4 ± 24.9 119.7 ± 25.8 0.098

Cerebellar ADCb 712.9 ± 70.6 730.3 ± 58.5 671.5 ± 81.7 0.028

Cerebellar ADC SD 139.4 135 149.8 0.016

rVolume 14.0 % ± 13.2 % 10.2 % ± 10.2 % 23.1 % ± 15.4 % 0.009

a (Cubic centimeters)b (910-6 mm2 s)

Neurocrit Care (2013) 19:276–282 279

123

Page 5: Prediction of Poor Outcome in Cerebellar Infarction by Diffusion MRI

A range of lesion volume (20, 25, 30, and 35 cc) and

rVolume (15, 20, 25, 30, and 35 %) cut-offs were tested for

specificity, sensitivity, PPV, NPV, TP, and FP. Results are

presented in Tables 3 and 4, respectively. A lesion volume

cut-off of 25 cubic centimeters (cc) identified subjects with

poor outcome with specificity (0.84), sensitivity (0.46),

PPV (0.55), NPV (0.79) with 6 TP and 5 FP. The cut-off of

30 cc had similar results with specificity (0.84), sensitivity

(0.38), PPV (0.5), NPV (0.76) with 5 TP and 5 FP. A

rVolume cut-off of 20 % identified subjects with poor

outcome at a specificity of 0.74, sensitivity (0.46), PPV

(0.43), NPV (0.77), with 6 TP and 8 FP. At a rVolume of

25 %, specificity improved (0.87) with sensitivity (0.31),

NP (0.5), NPV (0.75), and 4 TP and 4 FP. At the rVolume

cut-off of 30 %, specificity was (0.97), with sensitivity

(0.31), PPV (0.8), NPV (0.77), and 4 TP and 1 FP.

Discussion

We systematically examined potential MRI characteristics

of cerebellar infarction as possible predictors of poor out-

come in cerebellar stroke patients. Our definition of poor

outcome is similar to MCA lesion studies which include

death, secondary deterioration of consciousness with no

other obvious causes, and surgical interventions due to

swelling complication. We identified several quantitative

parameters that individually predicted poor outcome: DWI

lesion volume, whole cerebellar ADC, whole cerebellar

ADC standard deviation, and rVolume. rVolume and DWI

lesion volume retained significance when controlling for

age and NIHSS, the second and third most powerful pre-

dictors in the model, respectively.

In the anterior circulation, DWI can identify patients for

a malignant syndrome characterized by swelling and poor

outcome [17, 18]. The correlation between ADC mea-

surements and the presumptive underlying physiology,

severe ischemia, and decreased oxygen metabolism, lead-

ing to swelling and poor outcome has been prospectively

studied in MCA infarction but not in the posterior circu-

lation [18, 19]. This gap in knowledge and the early

identification of cerebellar infarctions that may swell is of

special clinical importance, precisely because neurological

deficits in this region often underestimate real infarct vol-

ume [30].

Fig. 2 a ROC curve for logistic

regression models for poor

outcome and b the generalized

linear model fitted values at a

range of sensitivities

Table 3 Results of specificity, sensitivity, positive predictive value

(PPV), negative predictive value (NPV), true positives (TP), and false

positives (FP) at a range of DWI lesion volumes in cubic centimeters

(cc)

Absolute volume (cc)

20 25 30 35

Specificity 0.742 0.839 0.839 0.935

Sensitivity 0.770 0.462 0.385 0.308

PPV 0.556 0.545 0.500 0.667

NPV 0.885 0.788 0.765 0.763

TP 10 6 5 4

FP 8 5 5 2

Table 4 Results of specificity, sensitivity, positive predictive value

(PPV), negative predictive value (NPV), true positives (TP), and false

positives (FP) at a range of rVolumes of Infarct as a percentage of the

whole cerebellum

rVolume of infarct (%)

15 20 25 30 35

Specificity 0.677 0.743 0.872 0.968 1

Sensitivity 0.769 0.462 0.308 0.308 0.231

PPV 0.500 0.429 0.500 0.800 1

NPV 0.875 0.767 0.750 0.769 0.756

TP 10 6 4 4 3

FP 10 8 4 1 0

280 Neurocrit Care (2013) 19:276–282

123

Page 6: Prediction of Poor Outcome in Cerebellar Infarction by Diffusion MRI

Our study is the first to consider the prediction of poor

outcome at several values of DWI lesion volume and

rVolume in cases of cerebellar stroke. Given the results of

regression and ROC analysis, the optimal lesion volume to

predict poor outcome may lie between 25 and 30 cc. Both

DWI lesion volumes had high specificity 0.84. Similarly,

an rVolume value between 25 and 30 % may best predict

poor outcome with specificities of 0.87 and 0.97,

respectively.

Clinical parameters have previously been used to predict

deterioration in cerebellar infarct with mass effect. The

finding in this study that NIHSS was not different between

outcome groups may support the theory that NIHSS is not the

ideal neurological exam to evaluate cerebellar stroke [26, 31,

32]. A prior study showed that hydrocephalus, brainstem

deformity, and basal cistern compression may signal dete-

rioration in cerebellar infarct with mass effect while infarct

volume does not [2]. Similarly, another prior study con-

cluded that cerebellar lesion size did not correlate with

International Cooperative Ataxia Rating Scale Subscore

[21]. Our study is unique in that it strictly uses diffusion-

weighted MR image analysis to predict an acute clinical

outcome. In this study, we did not evaluate subjective

qualitative classifications such as hydrocephalus, basal cis-

tern compression, or the fourth ventricle compression.

There are limitations to our study. First, the retrospec-

tive nature of this study and the low relative frequencies of

cerebellar stroke cases available at a single hospital limit

stringent control of subject selection and thus further

analysis. The retrospective nature of this study limits

control of the time to scan after a cerebellar infarct (63.7 h;

median 41.1 h); however, the time to scan was not a sig-

nificant predictor of poor outcome. The difficulty in early

cerebellar stroke diagnosis may offer explanation as to why

the time elapsed before MRI is lengthy in some patients

[5, 30], and future studies will require systematic MRI

studies at consistent time intervals. Furthermore, several

subjects with cerebellar stroke had concurrent brainstem

infarction. This feature was not found to be a significant

predictor of poor outcome. No significant difference in

average ADC values was seen in this study nor were ADC

measurements significant in models to predict poor out-

come. The significance of CAD/Prior MI may be a marker

of baseline health, though it is difficult to establish an

association with this dataset. The significance seen in PICA

territory strokes may be due to their previously defined high

frequency [6]. Given these limitations, this study is unique

in that it strictly examines MRI scans of cerebellar stroke

patients with a protocol considered both by a radiologist and

neurologist (RG and KS). Recently, automated systems

have been developed to select patients for stroke therapy

[33, 34]. A similar tool may be useful for cerebellar stroke.

Magnetic resonance imaging continues to be superior to

CT in acute stroke imaging due to its ability to detect acute

ischemia, especially in the posterior fossa [14, 16].

Apparent diffusion coefficient maps, as calculated from

diffusion-weighted images, are more sensitive to acute

infarcts that traditional T2 MR and ADC maps have

stronger interrater agreement and more accurately repre-

sent final infarct size [35, 36]. These findings suggest

diffusion MRI as the most effective parameter for stroke

imaging in the posterior fossa [14].

In conclusion, our analysis of 44 cerebellar stroke sub-

jects shows that rVolume and DWI volume may be specific

predictors of poor outcome, even when controlling for age

and NIHSS score in a logistic regression model. The results

of this retrospective analysis may place the best cut-off

value for prediction of poor outcome at a rVolume between

25 and 30 % or a volume between 25 and 30 cc. Stroke

MRI and volumetric analysis can help in the selection of

patients for early intervention to prevent the onset of

clinical deterioration or to identify those patients who are

at highest risk of swelling complications. In studies con-

sidering the prediction of malignant cerebral artery

infarction by MRI, retrospective endpoints and results

similar to those of this study were used as the basis for

prospective, multicenter validation study [17–20]. The

prospective study confirmed the initial retrospective results

with greater accuracy and controlled subject selection [18].

Prospective observational data of cerebellar stroke patients

are urgently needed to validate these initial findings. Cer-

ebellar stroke MRI lesion volume may help in the selection

of patients for early intervention to improve clinical

outcome.

Acknowledgments Dr. Sheth is supported by research Grants from

the American Academy of Neurology, American Heart Association

(11CRP5480009), and collaborations with Remedy Pharmaceuticals.

Mr. Tchopev is supported by a research training grant from the

Howard Hughes Foundation. Dr. Gullapalli is supported by research

funding from the United States Department of Defense.

References

1. Hornig CR, Rust DS, Busse O, Jaub M, Laun A. Space-occupying

cerebellar infarction. Clinical course and prognosis. Stroke.

1994;25:372–4.

2. Koh MG, Phan TG, Atkinson JL, Wijdicks EF. Neuroimaging in

deteriorating patients with cerebellar infarcts and mass effect.

Stroke. 2000;31(9):2062–7.

3. Chen HJ, Lee TC, Wei CP. Treatment of cerebellar infarction by

decompressive suboccipital craniectomy. Stroke. 1992;23(7):

957–61.

4. Mathew P, Teasdale G, Bannan A, Oluoch-Olunya D. Neuro-

surgical management of cerebellar haematoma and infarct.

J Neurol Neurosurg Psychiatr. 1995;59(3):287–92.

Neurocrit Care (2013) 19:276–282 281

123

Page 7: Prediction of Poor Outcome in Cerebellar Infarction by Diffusion MRI

5. Savitz SI, Caplan LR, Edlow JA. Pitfalls in the diagnosis of

cerebellar infarction. Acad emerg med. 2007;14(1):63–8.

6. Tohgi H, Takahashi S, Chiba K, Hirata Y. Cerebellar infarction.

Clinical and neuroimaging analysis in 293 patients. The Tohoku

Cerebellar Infarction Study Group. Stroke. 1993;24(11):1697–701.

7. Caplan L. Posterior circulation ischemia: then, now, and tomor-

row : The Thomas Willis Lecture—2000. Stroke. 2000;31(8):

2011–23.

8. Battaglia F, Quartarone A, Ghilardi MF, et al. Unilateral cere-

bellar stroke disrupts movement preparation and motor imagery.

Clin neurophysiol. 2006;117(5):1009–16.

9. Kerber KA, Brown DL, Lisabeth LD, Smith MA, Morgenstern

LB. Stroke among patients with dizziness, vertigo, and imbalance

in the emergency department: a population-based study. Stroke.

2006;37(10):2484–7.

10. Pfefferkorn T, Eppinger U, Linn J, et al. Long-term outcome after

suboccipital decompressive craniectomy for malignant cerebellar

infarction. Stroke. 2009;40(9):3045–50.

11. Juttler E, Schweickert S, Ringleb PA, Huttner HB, Kohrmann M,

Aschoff A. Long-term outcome after surgical treatment for space-

occupying cerebellar infarction: experience in 56 patients. Stroke.

2009;40(9):3060–6.

12. Heros RC. Surgical treatment of cerebellar infarction. Stroke.

1992;23(7):937–8.

13. Jauss M, Krieger D, Hornig C, Schramm J, Busse O. Surgical and

medical management of patients with massive cerebellar infarc-

tions: results of the German-Austrian Cerebellar Infarction Study.

J neurol. 1999;246(4):257–64.

14. Hwang DY, Silva GS, Furie KL, Greer DM. Comparative sen-

sitivity of computed tomography vs. magnetic resonance imaging

for detecting acute posterior fossa infarct. J emergency med.

2012;42(5):559–65.

15. Mullins ME, Schaefer PW, Sorensen AG, et al. CT and con-

ventional and diffusion-weighted MR imaging in acute stroke:

study in 691 patients at presentation to the Emergency Depart-

ment. Radiology. 2002;224(2):353–60.

16. Chalela JA, Kidwell CS, Nentwich LM, et al. Magnetic resonance

imaging and computed tomography in emergency assessment of

patients with suspected acute stroke: a prospective comparison.

Lancet. 2007;369(9558):293–8.

17. Oppenheim C, Samson Y, Manaı R, et al. Prediction of malignant

middle cerebral artery infarction by diffusion-weighted imaging.

Stroke. 2000;31:2175–81.

18. Thomalla G, Hartmann F, Juettler E, et al. Prediction of malig-

nant middle cerebral artery infarction by magnetic resonance

imaging within 6 hours of symptom onset: A prospective multi-

center observational study. Ann neurol. 2010;68(4):435–45.

19. Thomalla GJ, Kucinski T, Schoder V, et al. Prediction of

malignant middle cerebral artery infarction by early perfusion-

and diffusion-weighted magnetic resonance imaging. Stroke.

2003;34(8):1892–9.

20. Arenillas JF. Prediction of early neurological deterioration using

diffusion-and perfusion-weighted imaging in hyperacute middle

cerebral artery ischemic stroke. Stroke. 2002;33(9):2197–205.

21. Deluca C, Moretto G, Di Matteo A, et al. Ataxia in posterior

circulation stroke: clinical-MRI correlations. J Neurol Sci.

2011;300(1–2):39–46.

22. Kase CS, Norrving B, Levine SR, et al. Cerebellar infarction.

Clinical and anatomic observations in 66 cases. Stroke. 1993;

24(1):76–83.

23. Tegos TJ, Kalodiki E, Daskalopoulou S-S, Nicolaides AN.

Stroke: epidemiology, clinical picture, and risk factors—Part I of

III. Angiology. 2000;51(10):793–808.

24. Ivamoto HS, Numoto M, Donaghy RM. Surgical decompression

for cerebral and cerebellar infarcts. Stroke. 1974;5(3):365–70.

25. Jensen MB, St Louis EK. Management of acute cerebellar stroke.

Arch Neurol. 2005;62(4):537–44.

26. Boone M, Chillon J, Garcia P-Y, et al. NIHSS and acute com-

plications after anterior and posterior circulation strokes. Ther

Clin Risk Manag. 2012;8:87–93.

27. Vahedi K, Hofmeijer J, Juettler E, et al. Early decompressive

surgery in malignant infarction of the middle cerebral artery: a

pooled analysis of three randomised controlled trials. Lancet

neurol. 2007;6(3):215–22.

28. DeLong ER, DeLong DM, Clarke-Pearson DL. Comparing the

areas under two or more correlated receiver operating character-

istic curves: a nonparametric approach. Biometrics. 1988;44(3):

837–45.

29. Robin X, Turck N, Hainard A, et al. pROC: an open-source

package for R and S+ to analyze and compare ROC curves. BMC

Bioinf. 2011;12(1):77.

30. Edlow JA, DE Newman-Toker, Savitz SI. Diagnosis and initial

management of cerebellar infarction. Lancet neurol. 2008;7(10):

951–64.

31. Linfante I, Llinas RH, Schlaug G, Chaves C, Warach S, Caplan

LR. Diffusion-weighted imaging and National Institutes of

Health Stroke Scale in the acute phase of posterior-circulation

stroke. Arch Neurol. 2001;58(4):621–8.

32. Yoo AJ, Barak ER, Copen WA, et al. Combining acute diffusion-

weighted imaging and mean transmit time lesion volumes with

National Institutes of Health Stroke Scale Score improves the

prediction of acute stroke outcome. Stroke. 2010;41(8):1728–35.

33. De Silva DA, Desmond P. Utility of automated MRI analysis

software (RAPID) to select patients for reperfusion therapy: a

pooled analysis of the EPITHET and DEFUSE studies. Stroke.

2012;42(6):1608–14.

34. Lansberg MG, Lee J, Christensen S, et al. RAPID automated

patient selection for reperfusion therapy: a pooled analysis of the

echoplanar imaging thrombolytic evaluation trial (EPITHET) and

the diffusion and perfusion imaging evaluation for understanding

stroke evolution (DEFUSE) Study. Stroke. 2011;42(6):1608–14.

35. Bratane BT, Bastan B, Fisher M, Bouley J, Henninger N.

Ischemic lesion volume determination on diffusion weighted

images vs. apparent diffusion coefficient maps. Brain Res.

2009;1279:182–8.

36. Huang L, Wong X, Liu S, Li G. The application of DWI and

ADC map in cerebral infarction. Int Soc Magn Reson Med.2001;9:1446.

282 Neurocrit Care (2013) 19:276–282

123