Quantitative EEG for Non-seizure Indications

Fig. 1

Diffuse subarachnoid hemorrhage in the interpeduncular cistern and ambient cistern (a), extending to the suprasellar cistern and Sylvian fissures (b), with associated intraventricular hemorrhage and left frontal intraparenchymal clot (c)

Fig. 2

CT angiography demonstrating an ACA aneurysm

Table 1

Hunt and Hess Scale [3]

Grade	Clinical exam
1	Asymptomatic, mild headache, slight nuchal rigidity
2	Moderate to severe headache, nuchal rigidity, no neurologic deficit other than cranial nerve palsy
3	Drowsiness, confusion, mild focal neurologic deficit
4	Stupor, moderate-severe hemiparesis
5	Coma, decerebrate posturing

Table 2

Fisher Scale [4]

Grade	CT imaging findings
1	No detectable subarachnoid blood
2	Subarachnoid hemorrhage less than 1 mm thick
3	Subarachnoid hemorrhage more than 1 mm thick
4	Subarachnoid hemorrhage of any thickness with intraventricular hemorrhage (IVH) or parenchymal extension

Table 3

World Federation of Neurological Surgeons Grading system [5]

Grade	Glasgow coma scale score	Motor deficit
1	15	Absent
2	13–14	Absent
3	13–14	Present
4	7–12	Present or absent
5	3–6	Present or absent

Delayed cerebral ischemia (DCI) is one of the most significant complications that occurs after aneurysmal SAH and can be seen in up to 50 % of patients [6, 7]. DCI is often associated with radiographic vasospasm (Fig. 3) and has traditionally been considered the most probable cause of DCI. However, DCI can occur in the absence of vasospasm; hence, the two need to be distinguished [8]. The various pathophysiological mechanisms that have been proposed for delayed ischemia include vasospasm, cortical spreading depression, and loss of cerebral autoregulation [6].

Fig. 3

Right A1 vasospasm before (a) and after (b) treatment with intra-arterial nicardipine

Operationally, the diagnosis of delayed neurological decline due to ischemia usually distinguishes between imaging-confirmed cerebral infarction, and deficits attributable to ischemia in the absence of imaging confirmation, usually called “delayed ischemic neurologic decline” (DIND). More detailed definitions adapted from consensus definitions [9] are provided in Table 4.

Table 4

Consensus definitions of DCI and DIND

Delayed ischemic neurologic decline (DIND)

One of these:

A. New focal neurological impairment (i.e., hemiparesis, aphasia, apraxia, hemianopia)

B. Decrease of at least 2 points on the Glasgow Coma Scale

And all of these:

Must last at least 1 h

Must not be apparent immediately after aneurysm occlusion

Not attributable to other causes based on CT, MRI, or other laboratory studies

Delayed cerebral infarction (DCI)

One of these:

Cerebral infarction on CT or MR scan of the brain within 6 weeks after SAH

Cerebral infarction on the latest CT or MR scan made before death within 6 weeks

Cerebral infarction proven at autopsy

All of these must be true:

Not present on the CT/MRI within 48 h after early aneurysm occlusion

Not attributable to other causes such as surgical clipping or endovascular treatment. Hypodensities on CT attributable to ventricular catheter placement or intraparenchymal hematoma should not be counted as DCI

Adapted from Vergouwen et al. [9]

DCI/DIND is typically seen 4–12 days after the initial hemorrhage and is a major cause of morbidity and death [7, 10]. Risk factors for DCI/DIND include high clot burden in the basal cistern and thick ventricular clots, along with poor grade SAH [6]. Treatment is centered on ensuring adequate perfusion to the brain. This is accomplished by achieving euvolemia with fluid resuscitation. If symptoms of delayed ischemia persist despite euvolemia, the next step is inducing a state of hypertension using intravenous pressors and mineralocorticoids. The final step in managing DCI involves angiography, with intra-arterial injection of vasodilators, typically calcium channel blockers such as nicardipine, and angioplasty [11].

Given the morbidity and mortality associated with DCI, several screening modalities are used in neuro ICU to identify early signs of DCI. Transcranial Doppler (TCD) ultrasonography is one of the most commonly used screening modalities. TCD is used to measure blood flow velocity in the major cerebral arteries. For the anterior circulation, a mean blood flow velocity less than 120 cm/s is consistent with the absence of vasospasm, and a mean blood flow velocity greater than 200 cm/s is suggestive of cerebral vasospasm. Studies have shown TCD to have a sensitivity of 38–91 %, and a specificity of 83–100 % for detecting vasospasm [12–14]. However, TCD is operator dependent, and typically is only done once a day. Other diagnostic techniques used to screen for DCI and vasospasm include CT angiography, CT perfusion, xenon CT, and magnetic resonance imaging (MRI). CT angiography has been shown to have a sensitivity of 80 % and specificity of 93 % for the detection of vasospasm [15].

Continuous EEG Monitoring

Continuous EEG (cEEG) monitoring can be utilized for ischemia detection and is particularly attractive for detecting DCI. EEG has the advantage of providing continuous data, as opposed to TCD or radiographic data, and can serve as an important tool to detect ischemia prior to development of irreversible injury.

The impact of cerebral blood flow (CBF) on the EEG is shown in Fig. 4 [16]. CBF of 12–18 ml/100 g/min is effectively an ischemic threshold, and as CBF approaches this threshold, predominantly slower frequencies are seen on the EEG. As the CBF approaches the ischemic threshold, there is reversible cellular injury caused by decreasing adenosine triphosphate (ATP) and loss of the transmembrane potential [16]. As CBF decreases to less than 10–12 ml/100 mg/min, there is irreversible cellular damage and cell death, and the EEG reveals a suppressed pattern. Early detection of decreasing CBF creates an opportunity to treat ischemia prior to the development of irreversible injury or infarction, and hence makes cEEG a useful tool for early detection of DCI in SAH patients.

Fig. 4

Changes in EEG with decreasing cerebral blood flow [16] (Reprinted from [16] and BioMed Central is the publisher. RightsLink/Springer license no: 3562491446946)

SAH can produce several systematic changes in the EEG background, including slowing, periodic discharges, seizures, and impaired reactivity to external stimulation [16, 17]. cEEG patterns that have demonstrated predictive value for DCI include focal delta slowing corresponding with the area of injury, bursts of frontal biphasic delta waves, continuous rhythmic delta activity, and continuous polymorphic or unreactive delta [18]. In a retrospective study of high grade SAH patients, EEG changes suggestive of early ischemia were present before 78 % of DCI events [19].

Quantitative EEG Monitoring

The primary challenge in using cEEG to identify ischemia in real time is the time-consuming and subjective nature of raw EEG interpretation. qEEG monitoring provides an essential complementary set of tools to facilitate effective, sensitive, and timely detection of ischemia.

The key qEEG changes that signal the onset of ischemia are well illustrated in a spectrogram from an elderly medical ICU patient with sepsis who unexpectedly suffered a cardiac arrest while undergoing cEEG monitoring (Fig. 5). As evidenced in the figure, as cerebral ischemia ensues, there is early drop out of alpha frequencies followed by loss of delta frequencies and eventual suppression. The same changes are characteristic of ischemia in SAH patients with impending DCI, albeit with a progression that is typically much more gradual.

Fig. 5

Spectrogram changes in a patient with cardiac arrest. There is an initial loss of alpha frequencies (a), followed by loss of slower frequencies (b, c), and eventual suppression of the EEG (d)

Most clinical practice of qEEG for the early detection of ischemia is based on the observation that ischemia produces trends of decreased fast and increased slow oscillations in the EEG. Both studies are based on ratios of power within specific bands within the EEG spectrogram [15, 16].

Alpha-to-Delta Ratio

One method of detecting DCI with qEEG is based on the alpha-to-delta ratio (ADR) [16, 19]. The ADR is defined as the ratio of the sum of the power within two bands, an alpha band (8–13 Hz) and delta band (1–4 Hz), illustrated in Fig. 6. The ADR is often displayed either as a smooth curve, derived by applying a moving average to repeated sequential ADR measurements, or as a histogram showing sequential measurements from non-overlapping windows (Fig. 7). Significant or sustained decreases in ADR are considered “alarms” signaling impending DCI. In a study of qEEG in 34 high grade (Hunt and Hess (HH) 4 and 5) SAH patients, the ADR had the strongest association with DCI [16, 19]. Nine of 34 patients developed DCI and had a median decrease of ADR of 24 %. Among several possible rules for triggering an “alarm,” the study suggested two as having particular clinical utility. First, six consecutive recordings with a 10 % decrease in ADR from baseline had a sensitivity of 100 % and specificity of 76 % for subsequent DCI. Second, any single measurement with a 50 % ADR decrease had a sensitivity of 89 % and specificity of 84 % for subsequent DCI. Figure 7 shows an example of how the ADR ratio varies in relation to changes in GCS, neurological exam, imaging findings, and treatment [16, 19].

Fig. 6

Alpha-to-delta ratio (ADR). The ADR is defined as the ratio of the sum of the power within two bands, the “alpha” band (8–13 Hz) and delta band (1–4 Hz)

Fig. 7

Alpha-to-delta ratio calculated every 15 min and GCS score [19] (Reprinted from [19] with permission from Elsevier, License no: 3940350912437)

Relative Alpha Variability

Another common method used to assess DCI is relative alpha variability (RAV) as an early predictor of ischemia, operationally defined as vasospasm evidenced by TCD mean velocities of greater than 120 m/s in the middle cerebral artery (MCA) distribution and a Lindegaard ratio (MCA/internal carotid artery (ICA) velocity) of greater than 3, or evidence of vasospasm on conventional angiogram. This method was investigated in 32 SAH patients with HH1-HH3 grade hemorrhages [20]. This method assigns a score to histograms derived from 8- to 12-h segments of EEG. Bars in the histogram represent sequential measurements, derived from 2-min epochs, of the “alpha” to “total” power ratio, defined as the power within the 6–14-Hz (“alpha”) band expressed as a percentage of power within the 1–20-Hz band [16]. Periods with high variability are assigned a score of 4 (“excellent” RAV), whereas periods with nearly absent variability are assigned a score of 1 (“poor” RAV). Periods with intermediate degrees of variability are assigned a value of 2 (“fair” RAV) or 3 (“good” RAV) (Fig. 8). A deterioration of RAV by 1 visual grade in one or more monitored channels was considered to be an “alarm” signaling impending DCI. Inter-rater agreement using this visual scale was reported to be high, with 100 % agreement for cases with “excellent” and “poor” RAV, and 90 % agreement for cases with “good” and “fair” RAV [16].

Fig. 8

Visual grading scale for relative alpha variability (RAV) [20]. Each histogram is a time series of serially computed alpha-to-total power ratios (power in the 8–14-Hz band expressed as a percentage in the 8–20-Hz band). The time window shown is 8 h (Reprinted from [20] with permission from Elsevier, License no: 3940290199754 )

Of 19 patients with angiographic vasospasm, the RAV decreased by a mean of 2 grades, and improved as vasospasm improved. In ten of these patients, the RAV reduction was observed prior to the detection of angiographic vasospasm by a mean of 2.9 days (SD 1.73). Reduction in RAV had a positive predictive value of 76 % and negative predictive value of 100 % for vasospasm.

Limitations

A major historical limitation of qEEG is the effect of artifacts on its parameters. In current practice frequent manual inspection is required when using the ADR and RAV indices to avoid being misled. However, commercially available qEEG software has made steady improvements in automated artifact removal. Future studies should investigate whether these improvements allow more accurate and/or more efficient use of qEEG data for ischemia detection. Development of automated statistical trend detection algorithms currently under development are expected to further improve upon the present state of the art.

Practical Protocol

A practical protocol for using EEG/qEEG to monitor patients with SAH for early signs of ischemia is shown in Table 5. Clinical reporting of EEG for ischemia detection is generally more labor intensive than monitoring solely to detect seizures, as the findings of interest (new slowing, alpha attenuation, and asymmetry) are often subtle and develop gradually. Consequently, the authors recommend a disciplined, 4-step approach to reading and reporting cEEG in SAH, using the data layout shown in Figs. 9, 10, and 11: first, evaluate the raw EEG and accompanying spectrogram; second, inspect the ADR for any systematic downward trends; third, assign a visual RAV score to each vascular territory; and last, formulate an overall impression of the data, stating whether or not the findings suggest the development of ischemia.

Table 5

Protocol for continuous EEG monitoring in patients with SAH

Inclusion criteria for patients:

Nontraumatic aneurysmal SAH

Poor mental status (HH Grade IV,V) or IVH or thick cisternal blood (Fisher 3)

Purpose of EEG monitoring:

Ischemia detection

Seizure detection

Timing of monitoring:

Start within 2 days of admission, or STAT if indicated, and stop after 10 days of monitoring

EEG monitoring and reporting:

Continuous EEG monitoring; bedside raw EEG and quantitative EEG displayed