Acute-stage diffusion-weighted magnetic resonance imaging for predicting outcome of poor-grade aneurysmal subarachnoid hemorrhage

Abstract

We investigated the role of acute-stage diffusion-weighted images (DWIs) for predicting outcome of poor-grade subarachnoid hemorrhage (SAH). This study included 38 patients with poor-grade SAH who underwent DWI within 24 h after onset. DWI findings were divided into three groups on the basis of lesion area: none (N), spotty (S, ≦10 mm²), or areal (A, >10 mm²). We evaluated the correlation between preoperative DWI findings and clinical outcome, and the characteristics of DWI abnormalities. DWI abnormalities were revealed in 81.6% of cases (group S 34.2% group A 47.3%). All patients in groups N and S and 73.3% of patients in group A were treated radically. For those patients without rerupture, favorable outcomes were achieved in 100% of group N, 53.8% of group S, and 0% of group A. Abnormal lesions on initial DWI, which resulted in permanent lesions, showed a mean apparent diffusion coefficient ratio to the control value of 0.71, which was significantly lower than 0.95 observed in reversible lesions (P<0.01). We recommend radical treatment for even poor-grade SAH as long as the preoperative DWI shows no or only spotty lesions. DWI may provide an objective means to estimate the outcome of poor-grade SAH.

Introduction

Optimal treatment for patients with poor-grade aneurysmal subarachnoid hemorrhage (SAH) remains controversial, although radical treatment may confer a good outcome for some of these patients (Forgelholm et al, 1993; Inagawa et al, 1995; Le Roux et al, 1996; Suzuki et al, 2000; Ungersbock et al, 1994). Grading systems to define poor-grade SAH, such as the Hunt and Hess grade (Hunt and Hess, 1968), the Hunt and Kosnik grade (H&K grade) (Hunt and Kosnik, 1974), or the World Federation of Neurosurgical Societies (WFNS) Committee scale (Drake, 1998), sometimes fail to predict outcome, especially when patients are evaluated soon after bleeding (Bailes et al, 1990; Takagi et al, 1990). Their clinical manifestations may change rapidly after admission because of recovery processes after the initial rupture (Suzuki et al, 2000), development of brain edema and hydrocephalus, incidence of aneurysmal rerupture (Inagawa et al, 1995), and global brain hypoxia and hypoperfusion because of cardiopulmonary events associated with SAH (Hayashi et al, 2000). Sedation to prevent rerupture may further mask the changes in their neurological findings. An objective evaluation of pathophysiology in the acute stage is required to establish a treatment strategy and for predicting outcome in poor-grade SAH.

Magnetic resonance (MR) diffusion-weighted imaging (DWI) is a powerful tool for the noninvasive detection of early brain injury (Minematsu et al, 1992; Moseley et al, 1990) and provides higher sensitivity than computed tomographic scans for the evaluation of primary brain damage with acute SAH in experimental models (Busch et al, 1998; Jadhav et al, 2008) and in humans (Endo et al, 2005; Hadeishi et al, 2002; Weidauer et al, 2008). However, the diagnostic value of DWI for predicting outcome of poor-grade SAH has not been reported previously.

In this study, we sought to evaluate the role of DWI in predicting outcome in patients with poor-grade SAH. For this purpose, patients were divided into three groups—no lesion (N), spotty lesions (S), and areal lesions (A)—on the basis of their findings of DWI in the acute stage. We retrospectively compared the incidence of rerupture and symptomatic vasospasm and clinical outcome among these groups. We also examined apparent diffusion coefficients (ADCs) and followed-up findings of abnormal parenchymal lesions on initial DWI for a better understanding of the pathophysiology of such lesions.

Materials and methods

Patient Population

This study was based on a series of 356 consecutive patients with SAH due to aneurysmal rupture who were admitted to the Kohnan Hospital between January 2006 and December 2008. Among these patients, 77 showed poor-grade aneurysmal SAH (H&K grade IV or V). Of the 77 patients, 38 cases (18 cases with grade IV and 20 with grade V) fulfilled the entry criteria of H&K grade IV or V on admission, Fisher Group 3 (Fisher et al, 1980) (with or without intracranial hematoma), under 80 years old, and MR imaging that included DWI performed before angiography and operations within 24 h after symptom onset. Institutional review board approval and informed consent were obtained for all patients.

MR Imaging and Classifications

All MR investigations were conducted on a 1.5-T system (Signa imager, GE Medical Systems, Piscataway, NJ, USA) with a standard head coil. For DWI, a single-shot, echo-planar spin-echo sequence was used with the following parameters: matrix size, 128 × 128; field of view, 22 × 22 cm; repetition time, 6000 ms; and echo time, 78 ms with b values of 0 and 1000 secs mm⁻². Sixteen 6-mm-thick axial slices with an interslice gap of 2 mm, which included tissue from the entire supratentorial and infratentorial brain area, were imaged for 4 secs in each of the three gradient directions. Isotropic images were then constructed by averaging the three DWI. The T2-weighted images were obtained with the same slice location and thickness as DWI. Fast spin-echo axial sequence with the following parameters was used: repetition time 3000 secs, TE 83 ms, matrix 256 × 256, and field of view 22 × 22 cm.

We classified initial DWI high-intensity areas of the cerebral parenchyma into areal (more than 10 mm in diameter) and spotty lesions (10 mm or less in diameter). The patients were then divided into three groups: group A (n=18), single or multiple DWI lesions including at least one areal lesion (Figures 1A–1D); group S (n=13), single or multiple DWI lesions including only spotty lesions (Figures 1E–1H); and group N (n=7), no DWI lesions. Because a subarachnoid clot may itself show high intensity on DWI, parenchymal lesions were carefully selected by comparing DWI and computed tomography scan.

Figure 1.

Radiological findings on admission. (A–D) A representative patient in group A. (E–H) A representative patient in group S. (A and E) Computed tomography (CT) scans showing subarachnoid hemorrhage but no obvious lesions in the cerebral parenchyma. (B and F) Diffusion-weighted images showing abnormal high-intensity areal lesions in the right temporal lobe and the bilateral paramedian frontal lobes (arrows) in B and an abnormal high-intensity spotty lesion in the right paramedian frontal lobe (arrow) in F. Arrowheads indicate subarachnoid clots, referring to CT scan. (C and G) T2-weighted images showing no abnormal lesion in the brain parenchyma. (D and H) Apparent diffusion coefficient (ADC) maps of B and F, respectively, showing heterogeneous ADC reduction in the high-intensity lesions on DWI (arrows) in D, and decreased ADC in the spotty lesion (arrow) in H.

Treatment Strategy

The principal strategy during this study period was as follows. We performed radical treatment for aneurysmal obliteration in patients who improved to H&K grade III and IV within 72 h after onset. We also performed radical treatment immediately for those presenting with intracranial hematomas that showed significant mass effect. We chose supportive treatment for patients who had been in H&K grade V until 72 h after onset of SAH without symptomatic intracranial hematoma or hydrocephalus, or for patients with inappropriate general conditions for radical treatment. The radical treatment included neurosurgical clipping or endovascular embolization within 72 h after onset. For patients in the severe conditions, the endovascular treatment tended to be the procedure of choice over the neurosurgical clipping, whereas existence of intracranial hematoma usually indicated surgical clipping. Decompressive craniectomy and/or continuous cerebrospinal fluid drainage to counteract uncontrollable intracranial pressure were performed as supplementary procedures, if needed.

Common postoperative management was based on mildly hypervolemic therapy to prevent symptomatic vasospasm (Muench et al, 2007). If symptomatic vasospasm was not improved by induced hypertension therapy and if luminal narrowing was observed on angiography, angioplasty was performed immediately.

Cerebral infarction caused by cerebral vasospasm was defined as the appearance of a new, high-intensity area on DWI not caused by angiography or operation between 5 and 20 days after SAH.

The Glasgow outcome scale (good recovery, moderate disability, severe disability, persistent vegetative state, and death) was recorded at approximately 3 months after onset of SAH. Favorable outcome was defined as good recovery or moderate disability, and poor outcome was defined as severe disability, vegetative stage, or death.

ADC Calculation

To investigate the fate of DWI hyperintensity lesions, region of interest (ROI) analysis of ADC values was undertaken for 11 cases in group A and 11 cases in group S (Figure 2). A few cases with abnormal hyperintensity lesions on initial DWI were excluded from the ROI analysis because no follow-up MR scans with adequate quality were performed.

Figure 2.

Illustrative cases in ROI analysis for ADC values. (A–C) A patient in group A (Case A-2, see also Figure 3A and Table 4). (D–F) A patient in group S (Case S-10, see also Figure 3B and Table 4). (A and D) Diffusion-weighted images (DWI) on admission. (B and E) Apparent diffusion coefficient (ADC) maps of A and D, respectively. ROIs (white empty boxes, 5 mm² each) were placed to cover the entire abnormal high-intensity lesions on the initial DWI except for hematoma. To obtain the control ADC value of the patient, several white empty circles were placed on the apparently normal white matter in the frontal, temporal, parietal, and occipital lobes. (C and F) T2-weighted images in the chronic stage corresponding to A and D, respectively. Most of the abnormal lesions on the initial DWI corresponded to high-intensity lesions on T2-weighted image, but some DWI lesions resulted in apparently normal (iso-intense) finding. Asterisk indicates intracerebral hematoma cavities.

Diffusion-weighted image sequence data were processed using Functool software (Advantage Workstation version 4.4, GE Medical Systems) to obtain ADC values, which were calculated from diffusion trace images collected with b=0 and b=1000 on a pixel-by-pixel basis.

A control ADC value of each case was calculated as the mean of the mean ADC values for ROIs, which were placed on apparently normal frontal, temporal, parietal, and occipital lobes (Figures 2B and 2E, white empty circles).

Region of interests of 5 mm² in area were manually placed within the high-intensity lesions on DWI. Sufficient number of ROIs were placed to fill the entire lesion (Figures 2B and 2E, white empty squares). While locating ROIs, we paid special attention to avoid areas that were affected by susceptibility artifact, hematoma, and cerebrospinal fluid. To avoid locating ROIs in the sulci, we referred to the b=0 images in which the sulci appear as hyperintense signals. ADC values of each ROI were calculated as a mean±s.d. of ADC values for all pixels in the ROI. The ADC ratio for each ROI was then calculated as (ADC value of the ROIs)/(control ADC value).

By comparing initial DWI and chronic T2-weighted images, each ROI was judged as to whether it eventually fell into the T2-weighted image lesions or not. The ROIs that became high- or iso-intense on chronic T2-weighted images were separately analyzed for mean ADC values. The T2-weighted images in chronic stage obtained between 20 and 30 days after onset were compared with initial DWI (Figures 2C and 2F). It was not possible to coregister the T2-weighted images and DWI because there were often mass effect of hematoma and/or edema which may change over time. Instead, we identified each gyrus and sulcus in each image to match the anatomical regions as much as possible. High-intensity areas on T2-weighted images not attributable to initial hematomas, operative damage, or delayed spasm were considered to be late lesions related to initial SAH damage.

Statistical Analysis

Statistical analyses were performed using GraphPad Prism (Version 5; GraphPad Software Inc., La Jolla, CA, USA). We applied the χ²-test for the group comparison of categorical data. Numerical data in Tables 1, 2 and 3 were expressed as median (25 percentile, 75 percentile). Statistical comparisons between the groups were conducted using the Kruskal–Wallis test with Dunn's post hoc test for multiple comparisons. The ADC data were expressed as mean±s.d. Comparison of the ADC ratios between high and isotropic intensity lesions on the chronic T2-weighted image used the unpaired Student's t-test. Differences were considered significant at a P-value of less than 0.05.

Table 1. Baseline characteristics of the study population.

	N group, n=7	S group, n=13	A group, n=18
Age (years)	60.0 (52.0/73.0)	67.0 (55.0/75.5)	63.0 (53.5/72.5)
Male to female ratio	2:5 (28.6%)	4:9 (30.8%)	5:13 (27.8%)
H&K grade
IV	4 (57.1%)	10 (76.9%)	4 (22.2%)
V	3 (42.9%)	3 (23.1%)	14 (77.8%)
Location of aneurysms
Supratentorial	6 (86.0%)	11 (85.0%)	17 (94.0%)
Subtentorial	1 (14.0%)	2 (15.0%)	1 (6.0%)
Associated hematoma
Incidence	2 (28.6%)	6 (26.2%)	14 (77.8%)*
Volume (mL)	16.5 (6.0/27.0)	9.5 (1.6/17.4)	10.5 (5.0/31.2)
Ventricular dilatation	1 (14.3%)	10 (76.9%)**	6 (33.3%)***
Interval between onset and admission (h)	1.0 (0.6/5.0)	3.6 (0.8/5.9)	1.8 (1.0/2.6)

Abbreviations: H&K grade, Hunt and Kosnik grade.

Baseline characteristics of the study population. Numerical data were expressed as median (25 percentile/75 percentile). Statistically significant differences against corresponding values of group N were tested by χ²-test, ^*P<0.01, ^**P=0.021, and ^***P=0.024.

Table 2. Preoperative grade and surgical treatments.

	N group, n=7	S group, n=13	A groupa, n=18
H&K grade in the preoperative status
III	4 (57.1%)	6 (46.2%)	0b
IV	3 (42.9%)	6 (46.2%)	9 (50.0%)b
V	0	1 (7.6%)	9 (50.0%)b
Treatment of aneurysms
Clipping	3 (42.9%)	5 (38.5%)	4 (26.7%)
Coiling	3 (42.9%)	7 (53.8%)	6 (40.0%)
Trap/bypass	0	1 (7.7%)	1 (6.6%)
None (preoperative death by rerupture)	1 (1)	0	7 (3)
Supplementary procedures
Decompressive craniotomy	0	0	10 (55.6%)
CSF drainage	0	6 (46.2%)	6 (33.3%)

Abbreviations: CSF, cerebrospinal fluid; H&K, Hunt and Kosnik.

^aIn four cases treated conservatively, data in the preoprerative status represent those after the decision making of treatment.

^bStatistically significant differences as compared with corresponding values in group N (P=0.023) by χ²-test.

Table 3. Clinical course.

	N group, n=7	S group, n=13	A groupa, n=18
Rerupture (15.8%)	2 (28.6%)	1 (7.7%)	3 (16.7%)
Symptomatic vasospasm	2 (33.3%)	5 (41.7%)	4 (26.7%)
Outcomea
Good outcome (GR + MD)	5 (100%)	7 (53.8%)	0b
GR	3 (60.0%)	3 (25.0%)	0b
MD	2 (40.0%)	4 (33.3%)	0b
SD	0	5 (41.7%)	7 (46.6%)b
VD	0	0	4 (26.7%)b
D	0	0	4 (26.7%)b
Reasons of poor outcome
Primary brain damage	0	0	11 (73.4%)
Intraoperative complication	0	0	2 (13.3%)
Symptomatic vasospasm	0	4 (80%)	2 (13.3%)
Uncontrollable hydrocephalus	0	1 (20%)	0

Abbreviations: D, death; GR, good recovery; MD, moderate disability; SD, severe disability; VS, vegetative state.

^aCases with no rerupture.

^bStatistically significant differences as compared with corresponding values in group N (P<0.01) by χ²-test.

Results

In 31 of 38 cases (81.6%), abnormal findings in cerebral parenchyma were revealed on initial DWI. No abnormalities were found in seven cases. There were no abnormal lesions in the infratentorial intraparenchyma on initial DWI.

Clinical Results

Table 1 presents baseline characteristics of our study population. There were no significant differences among the three groups regarding age, sex, H&K grade at admission, and location of aneurysms. The initial computed tomography scan revealed associated intracranial hematomas (intracerebral, sylvian, and/or subdural) in 28.6% of patients in group N, which was significantly less than the 77.8% (P<0.01) found in group A, although there was no significant difference in hematoma volume. Ventricular dilatation was observed in 14.3% of patients in group N, which was significantly less than the 33.3% in group A (P=0.024), and 76.9% in group S (P=0.021). None of the angiograms obtained at admission revealed significant early vasospasm, or steno-occlusive lesions of arteries and veins (data not shown). There was no statistical difference among the groups in the time between onset of SAH and admission to our hospital.

When we evaluated the neurological status at the treatment decision for all cases (Table 2), 57.1% of patients in group N, 46.2% of patients in group S, and 0% of patients in group A recovered to H&K grade III within approximately 12 h after SAH onset. This improvement rate was significantly lower in patients in groups A than in those in groups N (P=0.023). All patients in group N, except for one case with rerupture, or all cases in group S, and 73.3% of patients in group A were treated radically with surgical or endovascular treatment, whereas 26.7% of patients in group A were managed supportively. Actually, 4 (26.7%) cases of group A, which were managed supportively, had been in H&K grade V until 72 h after onset. Continuous cerebrospinal fluid drainage was required in 46.2% of cases in group S and 33.3% of group A, while decompressive craniectomy was performed in 55.6% of cases of group A. Neither supplementary surgical procedure was required for any of the cases in group N.

Table 3 shows the incidence of rerupture, symptomatic vasospasm, and clinical outcome of patients at approximately 3 months after SAH onset. There were no significant differences in the incidence of rerupture and symptomatic vasospasm among the three groups. Two cases experienced rerupture postoperatively. One patient, in group N, underwent endovascular trapping for a ruptured vertebral artery dissection, but the dissection reruptured just after the procedure. The other patient, in group S, underwent the intra-aneurysmal embolization for a ruptured aneurysm in the posterior communicating artery but suffered from rerupture of the aneurysm 1 day after the procedure. The overall outcomes of cases without rerupture were evaluated at approximately 3 months. The proportion of good outcomes among patients in group N (100%) was significantly higher than that among patients in group A (0%, P<0.01), but there was no significant difference between groups N and S (53.8%). Most of group A patients (73.4%) experienced a poor outcome because of the primary brain damage revealed on initial DWI, whereas symptomatic vasospasm was a main factor (80%) among the reasons for poor outcome in group S.

Characteristics on MR Imaging

The fate of the initial DWI high-intensity lesions was investigated by ADC analysis in comparison with T2-weighted images in the chronic stage (Figure 2). The mean control ADC values were 776.7±91.4 × 10⁻⁶ mm² secs⁻¹ in group A, 770.9±68.6 × 10⁻⁶ mm² secs⁻¹ in group S, and 725.4±60.3 × 10⁻⁶ mm² secs⁻¹ in group N. No significant differences were observed among the groups (P=0.09 for groups A versus N and P=0.13 for Groups S versus N), although higher tendency was seen in groups A and S compared with group N. Areal lesions tended to locate superficially, for instance, in the lateral or medial side of the frontal lobe or of the temporal lobe. Most of the areal lesions appeared as a heterogeneous pattern on initial DWI (Figure 2A, see also Figure 1). In each patient, mean ADC ratios were calculated separately for ROIs that exhibited high- or iso-intensity on chronic T2-weighted image (Figures 2A–2C). These are summarized in Figure 3A and Table 4, which shows that in each case, the ADC ratios of ROIs that became high intensity were lower than those of ROIs that became iso-intensity. In areal lesions that became high-intensity lesions on the chronic T2-weighted image, the mean ADC ratio was 0.767±0.083, which was significantly lower than that of the areal lesions that became iso-intensity lesions (0.944±0.052, P<0.01) (Table 4).

Figure 3.

Graphs showing the mean apparent diffusion coefficient (ADC) ratios on admission in relation to resulting findings on the chronic T2-weighed images (A) group A (n=11) and (B) group S (n=11). The ADC ratio was calculated for each regions of interest (ROI) as the mean ADC value of the ROI divided by the control ADC value of the patient obtained from apparently normal white matter regions. Black squares indicate mean ADC ratios of the ROIs, which result in high-intensity regions on the chronic T2-weighted images (HIR, high-intensity region). Number of ROIs for each black square was between 5 to 282 depending on patients. White squares indicate mean ADC ratios of the ROIs, which resulted in the iso-intensity regions on the chronic T2-weighted images (IIR, iso-intensity region). Number of ROIs for each white square was between 5 and 230 depending on patients. Note that some patients had regions corresponding either HIR or IIR only. (C) Comparison of mean ADC ratios between regions that eventually fell into high- (black squares) or iso-intensity (white squares) regions on chronic T2-weighted images. All ADC ratios in A and B were plotted and averaged. The mean ADC ratio for regions that appeared as high–intensity regions on the chronic T2-weighted image (HIR, 0.713±0.022) is lower than that of regions appeared as iso-intensity in the chronic image (IIR, 0.947±0.011) (P<0.01). Error bars indicate standard deviations. Details of ADC value and ratio in each patient have been provided in Table 4.

Table 4. ADC values and ratios of DWI lesions.

Case no.	Location	Control ADC value	High-intensity region		Iso-intensity region
			ADC valuea	ADC ratio	ADC valueb	ADC ratio
Group A
A-1	FL, T	721.4	733.7±60.9	0.850±0.07	843.2±52.0	0.976±0.06
A-2	FL, T	863.6	611.2±62.6	0.701±0.07	773.6±49.6	0.888±0.06
A-3	FL, T	725.7	688.7±157	0.724±0.17	886.7±27.1	0.932±0.13
A-4	FL, T	951	616.3±81.1	0.683±0.09	787.8±53.1	0.872±0.06
A-5	FL, T	722.8	591.4±22.5	0.814±0.03	686.3±50.0	0.944±0.07
A-6	T	799.2	737.7±66.3	0.803±0.07	832.9±53.4	0.906±0.06
A-7	FM, T	870	749.3±13.0	0.861±0.01	861.7±53.4	0.990±0.06
A-8	FM, C	777.7	452.8±67.6	0.630±0.09	648.3±80.0	0.902±0.11
A-9	T	758.6	(−)	(−)	753.2±44.7	0.993±0.06
A-10	FM, S	615.2	446.8±103	0.726±0.17	628.6±61.3	1.021±0.10
A-11	FM	738.1	644.8±19.7	0.874±0.03	712.4±57.1	0.965±0.08
Mean		776.7±91.4	629.3±106	0.767±0.08	765.0±87.6	0.944±0.05
Group B
S-1	BG	907.3	640.1±63.2	0.708±0.07	(−)	(−)
S-2	FM, C	784.3	536.4±47.3	0.683±0.06	(−)	(−)
S-3	FL, O	749.5	527.3±45.1	0.703±0.06	(−)	(−)
S-4	PV	835.4	(−)	(−)	816.2±70.6	0.977±0.08
S-5	FM, FL, PV	846.8	592.2±8.70	0.699±0.01	756.9±29.0	0.894±0.03
S-6	PV	704.2	(−)	(−)	695.3±2.10	0.967±0.00
S-7	C,O	741	429.6±24.0	0.580±0.05	728.6±46.7	0.983±0.06
S-8	C	789.8	468.8±60.9	0.593±0.08	(−)	(−)
S-9	PV	704	453.4	0.644	(−)	(−)
S-10	S	702.4	372.3±93.5	0.53±0.13	(−)	(−)
S-11	PV	715.7	527.2±44.5	0.737±0.06	(−)	(−)
Mean		770.9±68.6	505.3±83.0	0.653±0.07	721.2±76.8	0.926±0.07

Abbreviations: ADC, apparent diffusion coefficient; BG, basal ganglial; C, cingrate gyrus; DWI, diffusion-weighted imaging; FL, lateral side in the frontal lobe; FM, medial side in the frontal lobe; O, occipital lobe; S, splenium; PV, paraventricle; T, temporal lobe.

Control ADC: a mean ADC values of apparently normal white matter. Other data are expressed as mean±s.d. ( × 10⁻⁶ mm² secs^-1).

^aMean ADC values in abnormal DWI regions on adnormal DWI regions on admission, which resulted in the high-intensity regions on the chronic T2-weighted images.

^bMean ADC value in abnormal DWI regions on admission, which resulted in the iso-intensiy regions on the chronic T2-weighted image.

Almost spotty lesions appeared to be homogenous on DWI in the acute stage and also on the chronic T2-weighted image (Figures 2D–2F). In contrast to the areal lesions, the spotty lesions tended to locate in the deep white matter or in the paraventricle region. Most of the spotty lesions exhibited high intensity on the chronic T2-weighted image (Figure 3B). The mean ADC ratio of spotty lesions that resulted in high intensity on the chronic T2-weighted image (0.653±0.072) was significantly lower than that result in iso-intensity (0.926±0.076, P<0.01) (Table 4). All together, the mean ADC ratio (0.713±0.022) in the regions that eventually fell into high-intensity regions was significantly lower than that of the regions that appeared as iso-intensity on the chronic T2-weighted image (0.947±0.011, P<0.01, Figure 3C).

Discussion

This study indicates, for the first time, that the outcome in poor-grade SAH can be predicted by the DWI performed within 24 h after symptom onset. This observation is derived from the following findings. For those patients who did not experience rerupture, favorable outcomes were achieved in 100% of the patients with normal finding in acute-stage DWI, which was significantly higher than that of the patients with the areal lesions (0%), while there was no statistical significance compared with those with the spotty lesions (53.8%). Furthermore, there was no significant difference in the incidence of rerupture, or symptomatic vasospasm among the groups.

We also examined the correlation between the reversibility of acute DWI findings and their ADC values to clarify the pathophysiology of such findings during the acute stage of poor-grade SAH. The mean ADC ratio in lesions that appeared as high intensity (probable permanent lesions) on chronic T2-weighted image was 0.713±0.022. This is lower than those in control regions, suggesting the predominant involvement of cytotoxic edema. The mean ADC ratio in the lesions that appeared as iso-intensity (probable tissue recovery) was 0.947±0.011. This is higher than in the former, suggesting a combination of cytotoxic and vasogenic edema.

Taken together, the DWI in the acute stage provides important information for predicting the clinical outcome of patients with poor-grade aneurysmal SAH as well as the destiny of the affected brain in the chronic stage.

Validity of Predicting Outcomes for Poor-Grade SAH on the Basis of DWI Findings in Acute Stage

More than 90% of untreated patients in poor-grade SAH are thought to die (Bailes et al, 1990), and natural history data and population-based studies indicate that only 5% of patients with grade V achieve favorable outcomes (Alvord et al, 1972; Bonita and Thomson, 1985). In this study, we separated poor-grade SAH cases into three groups on the basis of DWI findings in the acute stage. We managed radical treatment with aneurysm obliteration for all patients in groups N and S, and for 73.3% of those in group A, while providing supportive treatment for 26.7% of those in group A. In total, favorable outcomes were achieved in 31.6% of cases in this study. This result is consistent with previously published data (Bailes et al, 1990; Le Roux et al, 1996; Ungersbock et al, 1994).

In this study, the Glasgow outcome scale of the radically treated patients in group A was severe disability in seven, vegetative state in two, and death in two patients, whereas the Glasgow outcome scale of supportively treated patients in group A was vegetative state in two and death in five. These data suggest that the radical treatment implemented in group A may have increased the risk of surgical complications, and the risk of survival in poor condition. The primary brain damage revealed on initial DWI was a main factor influencing the outcome of the patients in group A. Conversely, spotty lesions on initial DWI appeared to have less significant impact on the outcome because symptomatic vasospasm was the main determinant of poor outcome in group S. In group S, where all patients were treated radically, favorable outcomes were achieved in 53.8% of cases. Thus, we recommend the radical treatment even for poor-grade SAH as long as the preoperative DWI shows no or only spotty lesions. Additional strategies such as decompressive craniectomy and/or mild hypothermia may be necessary and the matter of future investigation for the treatment of patients in group A.

Pathophysiology of Abnormal DWI Findings in the Acute Stage of SAH

A number of previous studies have reported abnormal DWI findings in acute SAH (Endo et al, 2005; Hadeishi et al, 2002; Jadhav et al, 2008; Liu et al, 2007; Shamoto et al, 2007; Shimizu et al, 2005; Weidauer et al, 2008). Hadeishi and his colleagues found that five of seven patients with acute poor-grade SAH (WFNS grades 4 and 5) had widespread multifocal patchy abnormalities throughout the brain, possibly caused by focal disturbances in cerebral perfusion throughout the brain due to dramatically increased intracranial pressure. Weidauer et al. (2008) investigated abnormal DWI patterns after aneurysmal SAH that included cortical band-like lesions adjacent to sulcal clots, which were distinct from territorial infarcts or laminar infarcts produced as a result of severe stenosis or occlusion in proximal vessels. They surmised that these band-like lesions were caused by a vasospastic reaction of the most distal superficial and intraparenchymal vessels.

In this study, we found two different patterns of abnormal lesions on DWI, areal lesions, and spotty lesions. The areal lesions that were more than 10 mm in diameter, located mainly in the cortical area, presented in a heterogeneous manner, and were often associated with intracerebral or sylvian hematomas in concordance with previous descriptions (Hadeishi et al, 2002; Weidauer et al, 2008). Thus, we speculated that some of these lesions were possibly caused by rapid and transient disturbances in the focal cerebral perfusion due to locally increased intracranial pressure and/or vasospastic reaction of distal superficial vessels. Conversely, the spotty lesions that were 10 mm or less in diameter and located mainly in the medial sides of the brain (including the frontomedial area, paraventricle, cingulate gyrus, and splenium) generally presented in a homogeneous manner and were frequently associated with ventricular dilatation. Taking into consideration that some cases with spotty lesions experienced a long delay between onset of SAH and admission, it is possible that some spotty lesions, especially those formed around the ventricles, may have resulted from gradually increased intraventricular pressure with diffuse hypoperfusion (Grote and Hassler, 1988) or hypoxia (Siskas et al, 2003). Microembolisms in perforators concomitant with aneurysmal rupture may also have contributed to the formation of other spotty lesions, especially in the deep white matter (Romano et al, 2002).

Besides these mechanisms, many investigators previously reported that other possible factors contribute to primary brain damage in acute SAH, including acute vasospasm (Taneda et al, 1990), vasogenic edema after early reperfusion (Nornes and Magnaes, 1972), and toxic effects of hemolytic cell products (Macdonald and Weir, 2001). Most of these factors can progress both temporally and spatially in the brains of poor-grade SAH patients at the acute stage, so the primary brain damage revealed on initial DWI is considered to result from the sum of such factors. Taken together, DWI findings may provide more information on extent and degree of the primary brain damage in patients with poor-grade SAH.

Characterization of ADC Values

For determining the fate of high-intensity intraparenchymal lesions on initial DWI in patients with poor-grade SAH, we investigated ADC values of their lesions and compared between initial DWI and T2-weighted image findings in the chronic stage. ADC, as a quantitative parameter of DWI, characterizes water mobility in local tissues for a variety of injury models (Moseley et al, 1990). High-intensity lesions on DWI with low ADC values are common in cytotoxic edema and are usually irreversible, although they can sometimes be reversed if early reperfusion is achieved (Fiehler et al, 2004). High-intensity lesions on DWI with normal or high ADC values may indicate vasogenic edema, which is usually reversible (Watanabe et al, 2002). However, there is little previous ADC data on intraparenchymal lesions in patients with SAH.

We revealed that DWI lesions in poor-grade SAH patients comprised a mixture of various ADC values, from normal to subnormal. We further showed that calculating ROIs/control ratios of ADC values yielded the threshold of the reversibility of such abnormal lesions on DWI at the acute stage of SAH. These values were consistent with the previous study using an experimental ischemic model (Sakoh et al, 2001) indicating that the lesions in which the ADC ratio was 0.713±0.022 and lower and that still presented as high-intensity regions on the chronic T2-weighted images contained more cytotoxic edema than vasogenic edema. The values also indicated that the lesions for which the ADC ratio was 0.947±0.011 and over resolved nearly or partially in the chronic stage and contained mainly vasogenic edema. In addition, the ADC ratios of the parts of the areal lesions that were revealed as high-intensity regions on the chronic T2-weighted image were higher than those of the spotty lesions (0.767±0.08 versus 0.653±0.07), probably because these parts of the areal lesions contained cytogenic edema intermingled with vasogenic edema. However, the ADC threshold for ischemic tissue viability must be estimated on the basis of the combination of time and degree, because the viability of injured brain tissue incorporates several factors, such as severity, duration, and volume of the insult (Sakoh et al, 2001; Miyabe et al, 1996; Heiss et al, 1998). In our series, there were some differences among groups in the duration between the onset and the admission, although all ADC values were calculated on the basis of DWI performed within 24 h after the onset of SAH. Taking the effect of time from onset on the observed ADC value into consideration, the ADC values may not be completely comparable between the groups in this study. Nevertheless, the evaluation of ADC values in poor-grade SAH patients at the acute stage may provide a more understanding of the distinctive characteristics of intraparenchymal lesions and also the prediction about the fate of such lesions at the chronic stage.

In addition, there was no significant difference among the mean control ADC values of the three groups in this study, ranging from 615.2 to 951.0 × 10⁻⁶ mm² secs⁻¹, but we did note a tendency for the values in group A to be higher than in group S and even higher than in group N. Some possible reasons for this diversity in the control ADC values are as follows. In patients with SAH, the ADC values in normal-appearing regions can fluctuate because of global mild vasogenic edema (Liu et al, 2007). It is possible that some pathological consequences causing vasogenic edema may exist in the apparently normal cerebral parenchyma in group A and, to a lesser extent, in group S. This idea may be supported by the observation that symptomatic vasospasm influenced on the outcome in group S more than in group N, although there is no significant difference in the incidence of symptomatic vasospasm. Further study in this regard is required for determining that a pathophysiology peculiar to poor-grade SAH may progress in regions not detected by DWI.

Study Limitation

Limitations of this study include the small number of patients, subjective comparison between initial DWI and chronic T2-weighted image findings, and the retrospective approach. However, this study may indicate that a prospective study is worth conducting.

Conclusion

In conclusion, DWI provides objective information regarding initial brain damage in poor-grade SAH. Abnormalities on DWI with the parenchymal involvement in the acute stage could be related to the clinical outcome of patients with poor-grade SAH. On the basis of our results, we recommend radical treatment even for poor-grade SAH as long as the preoperative DWI shows no or only spotty lesions. Future evaluation using a larger number of patients is warranted to address this important issue.

The authors have no personal financial or institutional interest in any of the drugs, materials, or devices described in this article.