Quantitative EEG monitoring for patients with subarachnoid hemorrhage

Electroencephalography and clinical Neurophysiology, 1991, 78:325-332 1991 Elsevier Scientific Publishers Ireland, Ltd. 0013-4649/91/$03.50 ADONIS 001346499100099F

325

EEG 90072

Quantitative EEG monitoring for patients with subarachnoid hemorrhage Douglas R. Labar a, Bruce J. Fisch a, Timothy A. Pedley a, Matthew E. Fink a

and Robert A. Solomon b Departments o/ " Neurology and h Neurosurgery, Neurological Institute o/New York, Columbia-Presbyterian Medical Center, N e t York, N Y (U.S.A.) (Accepted for publication: 13 August 1990)

Summary, We evaluated the sensitivity of continuous quantitative EEG in 11 patients with subarachnoid hemorrhage (SAH). We correlated compressed spectral array (CSA) and trend analysis (TA) of total power (1-30 Hz), frequency centroid (1-30 Hz), alpha ratio and percent delta power with clinical and radiological findings. For all ischemic events (n = 11), the most sensitive TA parameter was a change in total power (91%), followed by changes in alpha ratio (64%), frequency centroid (55%), and percent delta (45%). Comparable CSA features were changes in power (44%) and slowing (39%). Total power and frequency varied independently. In 4 cases, EEG findings on TA appeared before clinical changes. Continuous quantitative EEG may be useful for monitoring and predicting ischemia following SAH. TA of individual EEG parameters is more sensitive than CSA, and total power is the most sensitive. Key words: Subarachnoid hemorrhage; Quantitative EEG

Cerebral ischemia d u e to arterial vasospasm is a major source of morbidity in patients with aneurysmal subarachnoid hemorrhage (SAH), affecting 15-60% of patients (Solomon and Fink 1987). When threat of rerupture is eliminated by early surgery (Adams et al. 1987), delayed cerebral ischemia becomes the principal complication of concern. Early detection of the developing cerebral ischemia is particularly important because early therapeutic intervention using volume expansion (Pritz et al. 1978; Maroon and Nelson 1979; Solomon et al. 1984, 1988) induced hypertension (Kassell et al. 1982), nimodipine (Espinosa et al. 1984; Nosko et al. 1985), and high dose barbiturates (Kassell et al. 1980) may reduce mortality and morbidity (Solomon and Fink 1987). Routine and computerized quantitative EEG measures correlate with changes in cerebral metabolic function over a wide range of cerebral blood flow (CBF) values (Obrist et al. 1963; Sundt et al. 1974; Ingvar et al. 1976). In patients with cerebrovascular disease, quantitative computerized analysis correlates well with conventional EEG interpretation and, in addition, is more sensitive in some instances (Nagata et al. 1982, 1984: Sainio et al. 1983; Kopruner et al. 1984; Nuwer et

Correspondence to: Douglas R. Labar, M.D., Ph.D., Department of Neurology and Neuroscience, Division of Clinical Neurophysiology, New York Hospital-Cornell Medical Center, 525 East 68th Street, New York, NY 10021 (U.S.A.).

al. 1987). Although quantitative EEG analysis methods can now be implemented relatively easily for ICU monitoring (Karnaze et al. 1982), there have been few reports of its application to patients with subarachnoid hemorrhage (SAH). Using conventional EEG, Margerison et al. (1970) reported that regional vasospasm may produce localized EEG slowing. Parkes and James (1971) found that the extent of slowing detected by frequency analysis correlated with level of consciousness following SAH. Since EEG is a sensitive indicator of regional cerebral hypoperfusion, we considered that it might be useful in detecting onset of complications of SAH related to cerebral ischemia. To test this hypothesis, we examined the relationship between continuously acquired EEG spectral measures and the development of neurological complications due to presumed cerebral ischemia in patients with acute aneurysmal SAH.

Methods

We performed continuous quantitative 2-channel EEG monitoring in an unselected series of 11 patients admitted to the Neurological Intensive Care Unit with a diagnosis of SAH of presumed or proved aneurysmal origin. All patients had CT scans and 8 patients had o n e or more cerebral angiograms. Neurosurgical treatment for the aneurysm was performed b e f o r e or during the time of EEG monitoring in 5 patients (for further

326

clinical details, see Table I). Presumed ischemic events were determined on the basis of both clinical and radiological findings. Daily graded standardized neurological examinations were performed and recorded on all patients by the same examiner (D.R.L). The examination included the Glascow Coma Scale (Jennett and Teasdale 1982) and 24 other graded items to assess cortical and brain-stem neural function. Initial and subsequent angiogram and CT scan interpretations were recorded, as were significant metabolic derangements and medication changes. Standard EEG recording derivations were C,-'I~ and C,-Ta. These derivations were chosen because (1) a symmetric referential montage reduced the likelihood of asymmetries from non-cerebral activity; (2) they were largely free from eye movement potentials; (3) Cz is less likely to record muscle artifact than other 10-20 system electrode sites: and (4) these pairs provide a good summary of the major portion of lateral cerebral convexity activity, including some representation of territory supplied by the middle cerebral and anterior cerebral arteries. Occasional clinical circumstances, as in acute postoperative patients with bandaged incision sites, or known posterior circulation aneurysms, required modification of lead placement, but left-right symmetry always was maintained. The analog signal from each channel of EEG was filtered using a 20 Hz low pass filter (18 d B / o c t a v e ) and a 0.5 Hz high pass filter (6 dB/octave). A H a m ming window was applied to each 2 sec EEG epoch, and the power spectra were derived from a fast Fourier transformation. Automatic artifact detection methods eliminated EEG signals in excess of the dynamic range of the amplifiers, zero-derivative signals, and signals containing excessive 60 Hz interference. We compared various methods for processing and displaying the power spectra data. In all 11 patients, a 2-channel compressed spectral array (CSA) of sequentially stacked 2 min averages was displayed and printed in real-time at the bedside (Karnaze et al. 1982). Additionally in the (chronologically) last 6 of these, the spectrum from each 2 sec epoch was sent via cable to the Quantitative EEG Analysis Laboratory for storage and subsequent analysis. EEG power spectra were acquired continuously during the monitoring (i.e., 24 h/day): stored data files length varied depending on patient factors, such as transportation off the floor for radiological procedures (see Fig. 1). Trend analysis (TA) of stored quantitative EEG data from the 6 patients was performed by plotting selected spectral parameters vs. time. Each data point plotted represented an average of artifact-free 2 sec epochs collected over a 2 min period. An estimated > 95% of 2 rain intervals had more than 1 min of artifact free data. The quantitative EEG parameters analyzed included: (1) sum of the power (total power), 1-30 Hz; (2) centroid

D.R.

LABAR

ET AL.

of the frequency, 1 30 Hz (Cotas et al. 1979)~; (3) power 7.5-15 H z / p o w e r 1-7 Hz ('alpha ratio'): and (4) power 1-3.5 H z / p o w e r 1 30 Hz ('percent delta'). Trend analysis figures were created by plotting the contents of stored data files vs. time. Since stored data files varied in length, the resulting figures compared data files of different durations. EEG monitoring was continuous, 24 h / d a y , over the entire time each patient was studied, but it was necessary for data to be broken down into shorter subcomponent files for storage purposes. Since every 2 min an average power spectrum was stored, other parameters could have been derived. These 4 were selected prospectively and applied in a pilot fashion to test feasibility of this type of monitoring to this clinical situation. "I-A plots were compared initially using triangular weighted running mean and running median smoothing techniques. Because the triangular weighted running mean produced greater smoothing with better preservation of the data, we used this method exclusively for subsequent studies. We defined and prospectively identified noteworthy clinical events as: (1) presence or development of a focal neurological deficit attributed to ischemia or hemorrhage; (2) development of (or rarely recovery from) clinically obvious diffuse cerebral hemispheral dysfunction attributed to ischemia (i.e., global cortical dysfunction): and (3) encephalopathy at the onset of monitoring, seen immediately after the acute hemorrhage. Systemic metabolic parameters and medication effects were recorded and excluded as a cause of 'ischemic' events. We further required a focal neurological deficit to produce a score reduction by 1 / 3 1 / 2 the total scale for that item for a period of at least 24 h. Similarly, to be classified as a diffuse ischemic deficit, we required at least a 2-point reduction in neurological function on the Glascow Coma Scale for a period of at least 24 h in the absence of any non-ischemic cause. Plots of total power (1 30 Hz) and frequency centroid (1-30 Hz) vs. time were derived for all stored quantitative EEG data. We prospectively identified consistent EEG trends by visual inspection of these plots when clear day-to-day, or left-to-right, differences between groups of data persisted for more than 24 h. Since trends were required to persist for more than 24 h, it is unlikely they could be attributed simply to sustained changes in state of this duration. Consistent EEG trends in total power and frequency content were identified by CSA changes lasting more than 24 h. Visual inspection

30 H z

(F(f) × f) 1 Cenlroid

f - 0.5 30 H z

F~ v(r) 0.5

Q U A N T I T A T I V E EEG IN S U B A R A C H N O I D H E M O R R H A G E

of data display plots, rather than quantitative statistical comparison of data files, was carried out in order to minimize statistically significant, but not 'neurologically significant,' positive results (for discussion, see Oken and Chiappa 1986). Consistent EEG trends were characterized by less than 50% overlap between data plots compared (for example, see Fig. 1A). When there was more than 50% overlap, no consistent EEG trend was considered to be present (for example, see Fig. 1D). EEG analysis and identification of consistent EEG lrends for sum of the power, centroid of the frequency, and compressed spectral array were carried out prospeclively with the investigator blinded to the clinical events. After analyzing the quantitative EEG data, we compared clinical events, radiological findings, and quantitative EEG results, and determined sensitivities for the EEG findings. We also determined whether consistent EEG trends began prior to detection of any clinical neurological deficits. Specificity was evaluated by determining if consistent CSA or TA trends occurred in absence of clinical or radiological changes. Our analysis ,~f specificity did not include alpha ratio or percent delta. These 2 parameters were analyzed retrospectively only at times of clinical or radiological events, thus precluding recording 'false positive' EEG findings and calculation of specificities.

327

Results

Table I summarizes patient characteristics. Eight patients had encephalopathies with onset at time of initial SAH. Seven episodes in five patients of clinically detected diffuse ischemia attributable to diffuse vasospasm occurred. CT documented 3 cerebral infarctions (1 symptomatic, 2 asymptomatic) and 2 parenchymal hemorrhages (1 symptomatic, 1 asymptomatic). Cerebral angiograms revealed focal vasospasm in 2 of 8 patients. Both patients were symptomatic. In one of these, diffuse vasospasm also was present. Three patients with symptomatic ischemia did not show anglographic evidence of vasospasm, but the angiograms were not performed at the time of symptoms. Three other patients did not have ischemic symptoms or angiographic evidence of vasospasm. One patient with a symptomatic infarction on CT scan did not have anglographic evidence of vasospasm, but angiography was carried out well after the clinical stroke. TA and CSA were performed during 11 of these episodes of ischemia identified by neurological findings, CT, or angiography described above. An additional 7 of these described episodes occurred when only CSA was available. Consistent EEG trends using TA occurred in

FABLE 1 Patient characteristics. Age (years)

Clinical SAH grade (Hunt and Hess 1968) (28)

Angiogram aneursym location

Neurosurg. for aneurysm

Quant. EEG monitor method

Duration EEG monitored {days)

1 2 3 4 5

80 56 78 32 54

IV I Ill Ill 11I

No No No No Yes

CSA CSA CSA CSA TA CSA

4 4 10 1 10

6

69

[II

No angio. R PCA No angio. R ACA R ACA R PCA (R frontal AVM) No angio.

No

8

7 (first admit)

Yes

52

111

TA CSA TA CSA TA CSA TA CSA CSA

7

Patient no.

7 (second admit 8

52

lII

L L R R

PCOMA ANTCHORA PCOMA PCOMA

59

Ill

Normal

No

9

34

III

Yes

10

77

I

R ANTCHORA g ICA ( I N T R A C A VER.) Basilar

Yes

Yes

TA CSA

4

9 4 1

R = right, k = left, PCA = posterior cerebral artery, ACA = anterior cerebral artery, AVM = arteriovenous malformation, PCOMA = posterior communicating artery,, A N T C H O R A = anterior choroidal artery, ICA = internal carotid artery.

328

D.R. LABAR ET AL. A

20E-10

!8.0

~q

A o

3

/.

"

Day 3

Day4

0

3

6

9

0

Hours

3

6

9

Hours

80

3 O E-IO

S Z

D a y 4 PM . . ~

s

t~

!2

ooy4~MSl \

~oy4 A~,

,fit

Zf\ \dlS

i t..,..j

i x \_! \

-,.

l,..._l-~

F,

~, k../h-~

~, = , - . ,

/ "x,%; ~

,/

t\

v"--,,

i,

i

"

'li 0

3

6 Hours

9

0

3

6

9

Hours

Fig. 1. A: EEG trend analysis plots of total power (1-30 Hz) from a 69-year-old woman who was stuporous from a massive SAH. Between the 3rd and 4th days, she deteriorated from a confused stuporous state to one of semicoma. This was interpreted clinically as evidence of diffuse cerebral ischemia. Trend analysis plots showed reduced total power on day 4 (bottom line) relative to day 3 (top line) coincident with the clinical change (activity over left hemisphere illustrated, but EEG changes were bilateral). (Day 3 data file ends after 3 h of data collection, while day 4 data file ends after 9 h of data collection (see Methods)). B: trend analysis plots of frequency centroids showed slower frequencies on day 4 (bottom line) relative to day 3 (top line) coincident with the clinical change. The hour-to-hour fluctuations in the data seen in this and all other trend analysis plots are of uncertain significance, but may be related to patient state or transient fluctuations in intracranial pressure. C: EEG trend analysis showed improvement (recovery) in total power (1-30 Hz) on the night of day 4 (top line) compared with earlier in that day (bottom line). No clinical change was evident at the time power increased, but by day 5 the patient's level of consciousness also had improved and she responded to voice. Thus, EEG change of improved total power preceded the clinical recovery. D: trend analysis plots of frequency centroids during day and night of day 4 showed little difference between these time periods, but the trend analysis of total power (1-30 Hz) showed improvement, predicting the clinical improvement (see C). E: CT scan showed small left thalamic infarction which was clinically silent and present at start of monitoring. It became larger and better defined on subsequent CT scans, but remained asymptomatic. F: unsmoothed trend analysis plot of left vs. right difference in total power (1-30 Hz) revealed less total power over the left hemisphere throughout the study. This lateralization is consistent with the CT demonstrated thalamic infarct and is typical of the reduced power seen in 4 / 4 episodes of radiographically demonstrated focal ischemia. G: unsmoothed trend analysis plot of the left vs. right difference in frequency centroids over the same time period in this patient did not reveal a definite asymmetry.

QUANTITATIVE I::JE(} IN S U B A R A ( ' H N O I D H E M O R R H A G E

329

7 6 E-I ~

(+

7

I1,0E.1 0

1'1 5

10

15

Hours

c+~l 6 i, j il

.i,

i'

!

I~,[, I

1.3 0

5

!0

15

HOURS

Fig. 1 (continued).

11/11 ischemic events (overall sensitivity for T A = 100%; see Fig. 1). The range of degrees of change seen with these ischemic events is well illustrated by Fig. 1A, B, C, and F. Consistent EEG trends also occurred on CSA with 12/18 ischemic events (overall sensitivity for CSA =67%). For the 11 ischemic events monitored simultaneously with TA and CSA, consistent EEG

trends occurred in 6/11 (overall sensitivity of CSA in this subgroup = 55%). Thus, overall sensitivity of CSA for the 11 ischemic events monitored by both CSA and TA was similar to the overall sensitivity of CSA for all 18 ischemic events monitored (55% vs. 67%). CSA, but not TA, gave false negative results for 6 ischemic events (4 diffuse ischemia, 2 infarcts).

D.R. LABAR ET AL.

330 TABLE I1

TABLE III

Sensitivity of quantitative EEG parameters for various neurological events.

Quantitative EEG results associated with asymptomatic radiological findings.

EEG finding

EEG parameter

Focal CT lesions

All ischemic events (on exam, CT, angio.)

Asymptomatic radiological finding

Change in power

Trend analysis Compressed spectral array

100% (N = 5) 33% (N = 6)

91% (N = 11) 44% (N = 18)

Quantitative EEG results (ipsilateral to radiological finding)

Focal hemorrhage (CT)

Centroid (TA) Relative alpha (TA) Percent delta

60% (N = 5) 60% (N = 5)

55% (N = l l ) 64% (N = 11)

80% (N = 5)

45 % (N = 11)

17% (N = 6)

39% (N = 18)

1. Increased power (TA) 2. Slowing on centroids, relative alpha, percent delta (TA) 1. Increased power (TA) 2. Slowing on centroids, relative alpha, percent delta (TA) 1. Decreased power (TA) 1. Decreased power (TA) 2. Slowing on centroids, relative alpha, percent delta (TA) 1. Decreased power (TA) 2. Increased fast activity on centroids, relative alpha (TA)

Change in frequency

Focal hemorrhage (CT)

(TA) Compressed spectral array

Sensitivities of individual quantitative EEG parameters for detecting cerebral ischemia were calculated for the various clinical neurological and neuroradiological events (Table II). Change in power on trend analysis (91%) was the most sensitive measure for all types of presumed ischemic events (clinical, CT, angiographic). Alpha ratio (64%), frequency centroids (55%), percent delta (45%), total power on CSA (44%), and frequency on CSA (39%) were less reliable indicators. TA of total power show a 100% correlation with focal CT lesions, followed by percent delta (80%), frequency centroid and alpha ratio (60%), power on CSA (33%), and frequency on CSA (17%) (Table II and Fig. 1E-G). Five asymptomatic radiological abnormalities were accompanied by changes in total power on TA (Table III and Fig. 1 E - G ) , but none were detected by simultaneous CSA methods. No asymptomatic radiologic abnormality occurred without EEG change. In 4 cases, changes on quantitative EEG occurred before any change in clinical neurological status (Table IV and Fig. 1C). Radiographic studies revealed 4 instances of ischemia in 4 separate patients (1 symptomatic CT infarct, 2 asymptomatic CT infarcts (Fig. 1E-G), 1 asymptomatic focal angiographic vasospasm). Each was associated with an ipsilateral reduction of power on TA. In contrast, 2 focal hemorrhages (1 symptomatic, 1 asymptomatic) were associated with an ipsilaterat increase in power on TA. All incidents of reduced power were associated with presence or development of ischemia. Increased power occurred in 4 cases: 2 with focal hemorrhages, 1 with recovery from diffuse ischemia, and 1 with development of diffuse ischemia.

Thalamic infarction (CT) Subfrontal infarction (CT)

Focal vasospasm (angiogram)

Neurological examination revealed encephalopathic findings in 8 patients (6 monitored with TA + CSA, 2 monitored only with CSA). All occurred immediately after SAH onset and were not attributable metabolic derangement, medications, hydrocephalus, or postictal state. In all of these, there was bilateral slowing of EEG frequencies demonstrated by TA of the frequency centroid, alpha ratio, percent delta, and CSA. In 2 cases, EEG changes lasting more than 24 h occurred without detectable change in clinical or radiological findings. In 1 patient, unilateral slowing documented by TA frequency centroids, alpha ratio, and percent delta disappeared with a return to a symmetrical frequency pattern. This was not accompanied by any improvement in symptoms or change in the CT appearance of a subfrontal infarction. In the other patient, transient bilateral slowing was noted on CSA in the absence of any clinical or radiological change. A true specificity calculation could not be carried out,

TABLE IV Quantitative EEG predictive of clinical neurological events. Neurological event

No. of events

Quantitative EEG method predictive

Development of diffuse clinical ischemia Recovery from diffuse clinical ischemia Focal clinical ischemia

2

1. Increased power (CSA) 2. Slowing (CSA)

1

1. Increased power (TA)

1

1. Decreased power (TA) 2. Slowing on percent delta (TA) 1. Slowing on centroids, relative alpha, percent delta (TA)

Focal neurological deficit due to localized cerebral hemorrhage

331

QUANTITATIVE EEG IN SUBARACHNOID HEMORRHAGE

since no discrete unit for 'true negative events' could be defined.

Discussion

Our findings indicate that continuous quantitative EEG analysis may be a useful technique for monitoring the development of cerebral ischemia in patients with aneurysmal SAH. All ischemic events in our study patients were associated with consistent changes on TA (sensitivity = 100%). The corresponding sensitivity for CSA was 67%. Only 2 episodes of quantitative EEG changes (1 TA, 1 CSA) were not accompanied by new clinical or radiological findings. The sensitivity of TA was greater than CSA for power and frequency and for different types of neurological events. This was true even though the original power spectra data were the same for both methods. The superiority of TA over CSA was due to the greater ease with which asymmetries and variation in individual EEG parameters could be visualized using graphic plots. Total power was a more sensitive indicator of ischemia than other spectral measures (see Table II). Changes in overall amplitude (voltage attenuation) on routine EEG (a measure analogous to total power) are the best indicators of significantly impaired cerebral perfusion during carotid endarterectomy (Sharbrough 1987: Sundt et al. 1974; Trojaborg and Boysen 1973). In contrast, total power over a wide frequency band (in this study, 1-30 Hz) has not been a frequently employed quantitative EEG parameter in the study of stroke, lndeed, Sainio et al. (1983) reported better localizing value for various quantitative EEG frequency parameters other than total power (2-32.8 Hz). Substantial differences in patient population and methodology employed may account for these contrasting results. Calculation of specificity requires false positive and true negative values. Our data did not allow for a discrete 'true negative' unit. However, only 2 false positive EEG findings occurred in a total 62 days of monitoring, suggesting 'specificity' was good. Frequency and total power varied independently (e.g., see Fig. 1 and Tables Ill and IV), and total power sometimes changed without alteration in the relative distribution of activity within the frequency spectra. Furthermore, the relative contribution of various frequencies to the spectrum could change without apparent alteration in total power. Among patients with localized CT lesions, infarction produced reduced power with slowing on TA, whereas parenchymal hemorrhage produced increased power with slowing. These findings suggest that it may be possible to identify particular patterns indicative of specific causes of neurological dysfunction in patients with SAH. Although the electrophysiological accompaniments

of brain dysfunction in SAH patients are of theoretical interest, clinical application requires that this approach provide information not available with current routine techniques. In 4 cases, quantitative EEG changes preceded neurological deficits detected by daily neurological examination, and thus in these cases, the earliest indication of possible cerebral ischemia came from EEG. In 5 cases, quantitative EEG changes accompanied asymptomatic radiological findings, suggesting that EEG was superior to standard neurological examination for detecting cerebral dysfunction associated with these radiological abnormalities. We believe continuous EEG monitoring will become even more useful when practical methods for assessing wave form morphology are developed. We did not systematically compare 2-channel quantitative EEG monitoring with standard 20channel analog EEG, but the latter cannot be carried out without considerably more technical staff support and equipment than is needed for this quantitative method. Ease and speed of data review and analysis are enhanced by the continuous quantitative EEG monitoring method, particularly TA, in'comparison to visual analysis of original EEG from 20 channels derived from hundreds of hours of monitoring. However, potentially important data concerning wave form morphology and paroxysmal activity content are not adequately assessed by averaging spectral data. Moreover, improved methods of artifact detection are needed to insure the reliability of quantitative EEG analysis. We conclude that continuous quantitative EEG monitoring may have certain advantages (particularly in predictive value) over routinely employed methods of assessing CNS function in SAH patients, that continuous quantitative EEG monitoring accurately reflects neurological function, whether fluctuating or stable, and is useful for monitoring development of cerebral ischemia in patients with aneurysmal SAH.

References Adams, H.P., Kassell, N.F., Turner, J.C. and Haley, E.C. Predicting cerebral ischemia after aneurysmal subarachnoid hemorrhage: influences of clinical condition, CT results, and antifibrinolytic therapy. Neurology, 1987, 37: 1586-1591. Cotas, H.G., Bourne, J.R. and Teschan, P.E. Heuristic techniques in the quantification of the EEG in renal failure. Comp. Biomed. Res., 1979, 12: 299-312. Espinosa, F., Weir, B., Overton, T., Castor, W., Grace, M. and Boisvert, D. A randomized placebo-controlled double-blind trial of nimodipine after SAH in monkeys. J. Neurosurg., 1984, 60: 1167-1175. Heros, R.C., Zervas, N.T. and Varsos, V. Cerebral vasospasm after subarachnoid hemorrhage: an update. Ann. Neurol., 1983, 14: 599-608. Hunt, W.E. and Hess, R.M. Surgical risk as related to time of intervention in the repair of intracranial aneurysms. J. Neurosurg., 1968, 28: 14-20.

332 lngw~r, D.H., SjiSlund, B. and Ardo, A. Correlation between dominant EEG frequency, cerebral oxygen uptake, and blood flow. Electroenceph, clin. Neurophysiol., 1976, 41: 268-276. Jennett, B. and Teasdale, C. Management of Head Injuries. Davis, Philadelphia. PA, 1982: 78. Karnaze, D.S., Marshall, L.F. and Bickford, R.G. EEG monitoring of -clinical coma: the compressed spectral array. Neurology, 1982, 32: 117-124. Kassell, N.F., Peerless, S.J., Drake, C.G., Boarin, D.J. and Adams, H.P. Treatment of ischemic deficits from cerebral vasospasm with high dose barbiturate therapy. Neurosurgery, 1980, 7: 593-597. Kassell, N.F., Peerless, S.J., Drake, C.G., Boarin, D.J. and Adams. H.P. Treatment of ischemic deficits from vasospasm with intravascular volume expansion and induced arterial hypertension. Neurosurgery, 1982, I l: 337-343. Kopruner, V.. Pfurtscheller, G. and Auer, L.M. Quantitative EEG in normals and in patients with cerebral ischemia. In: G. Pfurtscheller, F.H. Lopes da Silva and E.J, Jonkman (Eds.), Brain Ischemia: Quantitative EEG and Imaging Techniques. Elsevier, Amsterdam, 1984:29 50. Margerison, J.H., Binnie, L.D. and McCaul, l.R. Electroencephalographic signs in the location of ruptured intracranial arterial aneurysms. Electroenceph. clin. Neurophysiol., 1970, 28: 296-306. Maroon, J. and Nelson, P.B. Hypovolemia in patients with subarachnoid hemorrhage: therapeutic implications. Neurosurgery, 1979, 4: 223-226. Nagata, K,, Mizukami. M.. Araki, G., Kawase, T. and Hirano, M. Topographic EEG study of cerebral infarction using computed mapping of the EEG. J. Cereb. Blood Flow Metab.. 1982, 2: 279-288. Nagata, K., Yonoki, K., Araki, G. and Mizukami, M. Topographic EEG study of transient ischemic attacks. Electroenceph. clin. Neurophysiol., 1984, 58: 291-301. Nosko. M., Weir, B., Kreuger, L. et al. Nimodipine and chronic vasospasm in monkeys. Part I. Clinical and radiological findings. Neurosurgery, 1985, 16: 129-135. Nuwer, M.R., Jordan, S.E. and Ahn, S.S. Evaluation of stroke using EEG frequency analysis and topographic mapping. Neurology, 1987, 37:1153 1159.

DR. LABAR ET AL. Obrist, W.D., Sokoloff, L., Lassen, N.A., Lane, M.H., Butler, R.N.B. and Feinberg, I. Relation of EEG to cerebral blood flow and metabolism in old age. Electroenceph. din. Neurophysiol., 1963, 15:610 619. Oken, B.S. and Chiappa, KH. Statistical issues concerning computerized analysis of brainwave topography. Ann. Neurol., 1986, 19: 493-494. Parkes, J.D. and James, I.M. Electroencephalographic and cerebral blood flow changes following spontaneous subarachnoicl hemorrhage. Brain, 1971, 94: 69-76. Pritz, M.B., Giannotta, S.E., Kindt, G.W., McGilicuddy, J.E. and Prager, R.L. Treatment of patients with neurological deficits associated with cerebral vasospasm by intravascular volume expansion. Neurosurgery, 1978, 3: 364-368. Sainio, K, Stenberg, D., Keskimaki. I., Mouroner, A. and Kaste, M. Visual and spectral EEG analysis in the ewduation of the outcome in patients with ischemic brain disease. Electroenceph. clin. Neurophysiol., 1983, 56: 117-124. Sharbrough, F. lntraoperative and intensive care electrophysiologic monitoring. In: O. Markand and R.P. Brenner (Eds.), 22rid Annual Course in Clinical EEG and Electrophysiotogy. American EEG Society, Atlanta, GA, 1987. Solomon, R.A. and Fink. M.D. Current strategies for the management of aneurysmal subarachnoid hemorrhage. Arch. Neurol., 1987.44: 769-774. Solomon, R.A, Post, K.D. and McMurtry, J.G. Depression of circulating blood volume in patients after subarachnoid hemorrhage. Implications for the management of symptomatic vasospasm. Neurosurgery, 1984, 15:354 361. Solomon, R.A., Fink, M.E. and Lennihan, L. Prophylactic volume expansion therapy for the prevention of delayed cerebral ischemia after early aneurysm surger. Arch. Neurol., 1988, 45: 325-336. Sundt, T.M., Sharbrough, F.W., Anderson, R.E. and Michenfelder, J.D. Cerebral blood flow and EEG during carotid endarterectomy. J. Neurosurg., 1974, 41:310 320. Trojaborg. W. and Boysen, G. Relation between EEG, regional cerebral blood flow, and internal carotid endarterectomy. Electroenceph, din. Neurophysiol., 1973, 34:61 69.