Electrophysiologic Monitoring in the Neurological Intensive Care Unit

Monitoring neurological functioning is of paramount importance in the neurological intensive care unit (neuro-ICU). However, even when performed conscientiously, serial neurological examinations are discontinuous, may miss important changes in the patient’s condition, and can vary according to the expertise of the examiner. Neurophysiologic monitoring with electroencephalography (EEG) and evoked potential (EP) are the primary tools available to the neurointensivist for objective evaluation of the functional status of the brain. In addition to supplementing the neurological examination as a measure of physiologic activity, EEG and EP can also provide unique diagnostic information related to specific brain wave patterns (e.g., burst-suppression, spindles, abnormal alpha, attenuation, seizures, and diurnal cycling on EEG studies) or anatomic function (e.g., evidence of brainstem damage on EP studies).

There are four primary applications for neurophysiologic monitoring in the neuro-ICU: (a) the detection of seizure activity and epileptic foci; (b) monitoring of neurological status (e.g., early detection of neurological deterioration owing to ischemia); (c) prognostication; and (d) evaluation of the effects of sedation and therapeutic interventions (1, 2, 3, 4, 5, 6, 7 and 8). In this rapidly evolving field, computer technology is being developed that can allow neurophysiologic monitoring to be fully integrated with other data collected in the neuro-ICU, such as cerebral perfusion pressure (CPP), intracranial pressure (ICP), and brain tissue oxygen tension. Multimodal monitoring of this type allows the neurointensivist to optimize physiologic variables that are directly under our control, such as CPP and PCO₂, and provide insights into how different physiologic derangements affect neurological function.

A single or even repeated electrophysiologic study may suffer from the same problems as does the neurological examination: It is a discrete, brief sample of data that may not reflect the patient’s overall condition. Many comatose patients, for example, show cycling of EEG patterns over the course of the day; these diurnal patterns cannot be detected by performing discrete electrophysiologic examinations. Seizures are usually brief and paroxysmal, and can be missed easily when EEGs are performed intermittently. For these reasons, monitoring of central nervous system (CNS) integrity or function in the neuro-ICU is best accomplished with continuous monitoring. Technologic advances have now made possible the collection, storage, analysis, and transmission of large amounts of continuous EEG and EP data (3). The ultimate goal of such monitoring is to enable the clinician to predict impending CNS injury at a time when intervention is still possible. For example, by the time temporal lobe herniation has reached the stage of third cranial nerve compression and pupillary dilation, a common clinical signpost, some irreparable damage probably has occurred already. In the case of cerebral vasospasm, hours can intervene between the onset of ischemia and the detection of clinical deficits on examination. Nonconvulsive status epilepticus (NCSE), left untreated, can exacerbate neuronal injury and become increasingly
refractory to treatment. These instances easily come to mind as examples in which continuous EEG (cEEG) monitoring might dramatically influence management in the neuro-ICU. Similarly, the reliable, objective data provided by EP studies can be useful sources of information about the level of functioning of specific CNS structures. This is especially true because short-latency somatosensory and brainstem auditory EPs are closely tied to anatomic structures, and are resistant to alterations by anything other than structural pathology. For example, brainstem auditory evoked potentials (BAEPs) and other subcortical EPs are essentially unchanged by high-dose barbiturate therapy sufficient to render the EEG isoelectric and clinical function virtually absent.

FIG. 8.1. Steps in generation of a compressed spectral array from segments of raw electroencephalographic (E.E.G.) data. (From Bickford RG. Newer methods of recording and analyzing EEGs. In: Klass DW, Daly DD, eds. Current practice of clinical electroencephalography. New York: Raven Press, 1979:451-480, with permission.)

ELECTROENCEPHALOGRAPHY

Background

As recently as ten years ago, EEG monitoring in most neuro-ICUs was performed intermittently with paper recordings. The use of digital cEEG monitoring initially was limited to dedicated epilepsy monitoring units, where the emphasis was on seizure classification and quantification, identification of interictal epileptiform discharges, and the presurgical evaluation of candidates for epilepsy surgery (9). Continuous electroencephalography monitoring has become increasingly accepted as a neuro-ICU monitoring technique with the advent of modern computer technology, which makes post hoc filtering, remontaging, adjusting of the sensitivity, and off-site reading
possible (10). Common indications for cEEG in the neurocritical care setting include the detection of nonconvulsive status epilepticus (NCSE), monitoring for cerebral ischemia and other causes of neurological deterioration, prognostication, and titration of sedation (best exemplified by titration of pentobarbital to attain burst suppression).

Computer processing of cEEG data also has the potential to reveal subtle changes over long periods of time that may not be evident when reviewing the raw EEG. Electroencephalography data collected over long periods of time can be transformed into power spectra by fast Fourier transformation (FFT), creating quantitative EEG (qEEG) parameters (Fig. 8.1). These can be displayed graphically as compressed spectral arrays (CSAs) (Fig. 8.2), and may reveal subtle changes in the EEG earlier than other monitoring techniques. With the advent of powerful microprocessors, data processing of this type can be performed in real time at the patient’s bedside. However, qEEG parameters of this type can be contaminated by artifact, and should not be attempted without proper training in electroecephalography and access to the raw EEG, to avoid the inclusion of signals related to artifact.

FIG. 8.2. Compressed spectral array of a normal adult, showing a large peak corresponding to the alpha rhythm. (From Bickford RG. Newer methods of recording and analyzing EEGs. In: Klass DW, Daly DD, eds. Current practice of clinical electroencephalography. New York: Raven Press, 1979:451-480, with permission.)

Technical Considerations

The practical challenges of cEEG recording in the neuro-ICU are substantial. The major hurdle is not owing to the complexity of the equipment or computer technology, but that of maintaining a low-impedance, lownoise connection between patient and machine. This challenge alone mandates special training of nursing staff and frequent daily visits by the EEG technician. In addition, there are multiple generators of electrical noise and artifact in the ICU, which can be unusually perplexing. Ventilators produce both mechanical and electric rhythmic artifacts. Nursing procedures such as chest physical therapy are another source of activity that can be in the frequency range of normal or pathologic brain electrical activity. Completely disconnected electrodes occasionally
may record activity that looks similar to a comatose patient with moderate to severe diffuse slowing and attenuation. Finally, sedating medications that can affect the EEG are routinely used in the neuro-ICU setting. For these reasons, the technologist, clinical neurophysiologist, ICU nurse, and neurointensivist must work closely together to ensure that the EEG accurately reflects cerebral activity. Team training and continuing education for clinicians are crucial to the success of any neuro-ICU cEEG monitoring program (5). For cEEG, valid interpretation is often only possible when either the ICU staff makes notes in the EEG record, or when a continuous video recording is combined with the EEG recording. Digital video has been shown to be particularly helpful for identifying subtle ictal phenomena (e.g., facial twitching or rhythmic eye movements) and EEG abnormalities related to artifact (12).

Specific Applications of Continuous Electroencephalographic Monitoring in the Neurological Intensive Care Unit

Detection of Seizure Activity

Indications

Patients in coma often experience nonconvulsive seizures that have few or no clinical manifestations. Continuous electroencephalographic monitoring is essential for the detection and prompt treatment of these events, which may contribute to the patient’s depressed level of consciousness or cause additional brain damage. Nonconvulsive seizures and nonconvulsive status epilepticus (NCSE) are common in all types of acute brain injury, and are not restricted to patients with epilepsy or those with a clinical diagnosis of seizures. In unselected neuro-ICU patients, cEEG monitoring reveals nonconvulsive seizures in up to 34%, and up to 75% of these cases have NCSE (13). Even after excluding patients with any clinical suspicion of seizures, cEEG detects NCSE in 5% to 10% of comatose medical intensive care unit (14), traumatic brain injury (TBI) (15), or subarachnoid hemorrhage (SAH) patients (16). Similarly, cEEG detects nonconvulsive seizures in up to 27% of patients with altered level of consciousness from any cause (17), in 36% of patients after the termination of generalized convulsive SE (18), in 22% of severe TBI patients (15), and in up to 23% of patients with intracerebral hemorrhage (19,20). It is important to diagnose NCSE in these patients, because the excessive metabolic demand of ictal activity may increase ICP and further compromise brain tissue at risk for ischemic or excitotoxic injury (8). The prognosis of NCSE in the ICU setting is poor: The overall mortality rate is 30% to 50% (16,21, 22 and 23), and medically refractory NCSE after severe TBI and SAH has been associated with 100% mortality (16,17). The challenge for the neurointensivist remains in demonstrating that the aggressive, early termination of NCSE in these patients can result in better outcomes.

In the absence of cEEG monitoring, appropriate treatment for patients with NCSE without overt clinical symptoms of seizures often is delayed. This may have a deleterious impact on outcome because delayed anticonvulsant therapy for SE has been associated with poor outcome (23,24) and an increased likelihood of refractory SE (25,26). Patients on continuous intravenous antiepileptic drugs (cIVAEDs) for the treatment of refractory SE always should be monitored with cEEG, because subclinical seizures may occur in up to half of patients during treatment, and up to 26% after the initial discontinuation of therapy (27). Although the outcome of NCSE is often poor, termination of ongoing electrographic seizure activity can result in recovery of consciousness and clinical improvement (25), and there remains little doubt that nonconvulsive seizures can directly injure the brain and cause enduring neurological impairment (28,29). In years past, “seizures, coma, and death” was an all-too-common course of events for victims of acute, severe brain injury. The advent of cEEG monitoring raises the possibility that untreated NCSE played a major role in prolonging coma and exacerbating brain injury in many of these cases.

Challenges of Electroencephalography Interpretation in Status Epilepticus

The accurate identification of electrographic seizures in comatose, severely brain injured patients presents special challenges for the electroencephalographer and neurointensivist (Table 8.1). Classic ictal patterns (Fig. 8.3) showing paroxysmal high-voltage spike and sharp-wave discharges (gleaned from epilepsy patients with otherwise normal brain function) may not be evident in the seizing patient with diffuse brain injury and profound EEG background suppression. Instead, more subtle discharges—such as paroxysmal waxing-and-waning focal slow wave activity or periodic lateralized epileptiform discharges (PLEDs)—may represent ictal activity in the comatose patient (Fig. 8.4) (30,31). In patients with known staus epilepticus, these EEG findings may be considered as part of an ictal/interictal continuum because they do not meet formal seizure criteria, and because their exact nature and significance remain poorly understood. Some have used serial EEG data (31,32), focal increased blood flow on single photon-emission computed tomography (SPECT) (33), and increased metabolism on fluorodeoxyglucose positron emission tomography (PET) scanning (34) to argue that PLEDs and other subtle EEG findings following SE can be ictal. Others regard PLEDs as a purely nonictal phenomenon. Further work in this controversial area is badly needed.

TABLE 8.1. Criteria for electrographic seizures

Repetitive spikes exceeding three per second

Repetitive spikes less than three per second, if changes after antiepileptic drug administration

Rhythmic waves with incrementing onset or decrementing offset, and postdischarge slowing or attenuation

From Jordan KG. Continuous EEG monitoring in the neuroscience intensive care unit and emergency department. J Clin Neurophysiol 1999;16:14-39, with permission.

Seizure Detection Software

Specialized EEG signal processing software can be used for screening large cEEG datasets for possible electrographic seizure activity, thus improving the efficiency of cEEG analysis. However, technical challenges remain. The data reduction techniques used in most cEEG CSA displays make recognition of spikes and sharp waves impossible, burst suppression almost impossible, and electrographic seizure activity difficult. Specific patterns related to the rhythmicity of electrographic seizure activity (35) or accompanying muscle artifact (36) may be helpful indicators. Tasker and associates (37) described continuous monitoring using the cerebral function analyzing monitor (CFAM) and found that paroxysmal events were clearly seen, although the sensitivity of this system for detecting partial seizures was limited. Another approach to solving the problem of seizure detection has been developed by Gotman, who published an algorithm for selecting parameters extracted from the digital EEG to identify ictal activity (38). Although the false-positive rate was quite high, the sensitivity was good, and few real ictal events are missed. Vespa and coworkers (15) have developed automated seizure detection software based on multichannel, digitized real-time FFT (2 seconds per epoch, 2 minutes average) and trends of total EEG power with some success.

Existing automated seizure detection software that has been developed for use in epilepsy monitoring units has largely been “trained” on seizures obtained from healthy, neurologically intact patients with seizure disorders. However, these seizures differ from those of comatose brain-injured patients, in whom ictal activity is often less organized, slower in maximum frequency, of longer duration, and without a clear on- and off-set. In order to use seizure detection programs in the neuro-ICU, new software needs to be developed that can more accurately identify potentially ictal patterns in patients with brain injury.

FIG. 8.3. Partial electrographic seizure with characteristic evolution (of frequency and amplitude) and offset. Top panel: The probable onset of the seizure is in the right parietal-occipital region (P4-O2 leads); prominent muscle artifact is seen in the other leads. It evolves into a run of high-frequency spike wave activity in the C4-P4 and P4-O2 leads. Bottom panel: The seizure evolves into slower frequency sharp wave activity. (Figure provided courtesy of Dr. Bryan Young.)

FIG. 8.4. Evolution of electroencephalographic findings in a 55-year-old woman with nonconvulsive status epilepticus after subarachnoid hemorrhage (selected montage of temporal electrodes only). Panel 1: Day 5 after subarachnoid hemorrhage: Mild to moderate diffuse background slowing, higher voltages on the right owing to skull defect from craniotomy, no epileptiform discharges. Panel 2: Day 5 (2 hours after panel 1): Broadly distributed right hemispheric periodic lateralized epileptiform discharges with underlying semirhythmic delta, diffuse background slowing, and left hemisphere attenuation. Panel 3: Day 5 (4 hours after panel 2): Bilateral seizure activity, right greater than left. Panel 4: Day 6 (24 hours after panel 3): Similar pattern to electroencephalograph in panel 2. (From Dennis LJ, Claassen J, Hirsch LJ, et al. Nonconvulsive status epilepticus after subarachnoid hemorrhage. Neurosurgery 2002;51;1136-1144, with permission.)

Neurological Monitoring

Electroencephalographic Signal Processing

The sensitivity of the EEG to changes in CBF and pharmacologic agents suggests that monitoring cEEG might bridge the gap between serial neurological examinations in the neuro-ICU setting, particularly in comatose patients. The primary challenge is to convert the complex EEG signal into a more “user-friendly” summary format that can be interpreted by nonelectroencephalographers in real time. The computer analysis technique most commonly used to accomplish this is transformation of the EEG into CSA. Compressed spectral array based on FFT of the EEG signal can be plotted as total power or overall amplitude, frequency activity totals (e.g., total or percent alpha power), spectral edge frequencies (e.g., the frequency below which 75% of the EEG record resides), and frequency ratios (e.g., alpha-delta ratio, ADR) (39, 40, 41 and 42). Other EEG data reduction display formats include the CFAM, the pEEG monitor, EEG density modulation, automated analysis of segmented EEG (AAS-EEG), and the bispectral index (BIS) monitor (6,43,44). Despite great promise, the utility of these compressed cEEG parameters for neurological monitoring in the neuro-ICU setting has yet to be demonstrated.

Cerebral Ischemia

Cerebral ischemia results in EEG changes because cortical layers 3 and 5, which are particularly sensitive to oxygen deficits, contribute most to the generation of electrical dipoles detected by EEG (8). It has long been known that infarction may result in polymorphic delta, loss of fast activity and sleep spindles, and focal attenuation (45,46). These EEG findings have been shown to reflect abnormal cerebral blood flow (CBF) and metabolism (cerebral metabolic rate of oxygen) as demonstrated by PET and Xenon-CT-CBF imaging (47,48). Electroencephalography is very sensitive for ischemia, and usually begins to change at the time reversible neuronal dysfunction occurs (CBF 25 to 30 mL/100 g per minute) (46), a level at which therapeutic interventions might be instituted to prevent permanent brain damage. On the reverse side, EEG is also very sensitive for recovery and may demonstrate recovery of brain function from reperfusion earlier than the clinical examination (8).

Intracranial Pressure

Relatively little work has explored the relationship between EEG patterns and ICP. In one study of 16 patients, Munari and Calbucci (49) found no single EEG pattern that correlated with mean ICP values. However, when the ICP was stable and without pressure waves, the EEG contained regular high-voltage slow waves, whereas an alternating EEG was seen in those patients with Lundberg B waves. Other human studies have also shown a poor correlation between ICP levels and EEG activity (50). It appears likely that EEG activity is not reliably influnced by increased ICP until very high levels are reached and CPP and CBF is compromised. In monkeys, Langfitt and associates (51) found that a gradual rise in ICP produced by an intracranial balloon produced no EEG changes until near levels approaching mean systemic arterial pressure were reached.

Intraoperative Monitoring

The relationship between EEG and ischemia was first applied in the early 1970s to intraoperative monitoring during carotid endarterectomy (52,53). Detailed studies of EEG parameters during carotid clamping show three patterns corresponding to increasingly severe changes in CBF. At the first level, conventional EEG traces remain unchanged to visual inspection, but computer analysis shows a 5% to 15% drop in amplitude without changes in peak or median power frequencies (52,53). This EEG pattern presumably corresponds to a minor drop in CBF. At the second level, the EEG shows marked slowing and an increase in amplitude. At the third and most critical level of ischemia, the EEG shows a marked loss of amplitude (greater than 50%) and further slowing of the dominant frequencies. This has been considered to be a significant change and an indication for a temporary bypass shunt if the clamp is to be maintained for more than a few minutes. Though the usefulness of EEG monitoring for carotid surgery is well established, the most sensitive parameters to monitor have not been clearly determined. However, it is established that computerized qEEG is more sensitive to change than routine visual analysis (54), and that the ADR is highly sensitive to the effects of ischemic stroke (55). Future innovations to improve the sensitivity and specificity of intraoperative monitoring may include automatied multiparameter EEG analysis using sophisticated methods such as wavelet analysis, fuzzy logic, and neural networks (56).

The techniques of interventional radiology also have spurred efforts to improve neurological monitoring, and have provided a rich source of information applicable to the neuro-ICU (Fig. 8.4). Hacke and coworkers (57) originally proposed that cEEG monitoring might be useful during angiographically guided balloon test occlusion (BTO) of the carotid artery. Several investigators have described EEG CSA slowing in response to BTO, which presumably reflects critical perfusion failure and CBF reduction (58,59). The
endovascular treatment of arteriovenous malformations (AVMs) has provided another opportunity for monitoring CNS function under highly controlled conditions. In one study, focal EEG slowing during preembolization superselective amytal testing was been shown in some series to predict subsequent neurological events (60). More recently, interventional neuroradiologists have described the use of intraarterial catheter tip EEG electrodes for the detection of seizure foci related to AVMs and deep brain structures not detected by conventional EEG (61,62).

Neurological Intensive Care Unit Monitoring for Delayed Ischemia After Subarachnoid Hemorrhage

The EEG changes observed during iatrogenic CBF changes in the operating room and angiography suite suggest that cEEG might be helpful in monitoring cortical function in the ICU. This application is best exemplified by SAH patients at risk for developing cerebral vasospasm (Fig. 8.5) (63, 64 and 65). Several studies have attempted to identify specific EEG parameters that correlate with the development of delayed cerebral ischemia after SAH, with varying results. Labar and colleagues (64) found that the trend analysis of total power (1 to 30 Hz) correlated with the development of delayed ischemia. Vespa and coworkers (63) identified the variability of relative alpha (6 to 14 Hz/1 to 20 Hz) as a predictor of vasospasm. Claassen and associates (65) found in a systematic analysis of 20 derived cEEG parameters that the poststimulation ADR (PSADR, 8 to 13 Hz/1 to 4 Hz) had the best sensitivity and specificity for delayed ischemia. Surprisingly, all of these studies found that focal ischemia sometimes resulted in global or bilateral changes in the EEG, and two found clear evidence that EEG changes can precede clinical deterioration by several days (63,64). Claassen and associates (65) studied poor-grade SAH patients (Hunt-Hess grade 4 or 5), in whom delayed ischemia and infarction is often not detectable by the clinical examination. They found that both a onetime 50% reduction or a prolonged greater than 10% decrease in PSADR from baseline were equally sensitive and specific (70% to 80%) for detecting delayed ischemia. Rivierez and coworkers (66) did not monitor EEG continuously, but obtained EEGs on days 1 and 5 after SAH in 151 patients. They found that a normal EEG correctly predicted a vasospasmfree course in 73%, but did not determine whether the EEG added value after considering clinical grade. Certain abnormal EEG patterns, especially one showing broad, repetitive slow waves, termed “axial bursts,” predicted clinical or angiographic evidence of vasospasm 97% of the time. Although cEEG monitoring for delayed ischemia is a promising application, additional research is needed to develop online real-time cEEG parameters continuously displayed in the neuro-ICU.

One drawback of quantitative EEG interpretation is that absolute values have to be interpreted cautiously because of intersubject differences. For this reason, cEEG is likely to be most useful for trending cerebral activity after a baseline assessment can be obtained. Within-subject state changes, variations in physiologic parameters (e.g., cerebral perfusion pressure), medications (i.e., sedatives), and artifacts can heavily influence qEEG parameters (67). For this reason, most studies have found that relative qEEG parameters are more useful than absolute parameters (63,65). These studies also have found that the best data are obtained when artifact-free clips of EEG are obtained following physical stimulation.

Other Neurological Intensive Care Unit Continuous Electroencephalography Monitoring Applications

A variety of additional applications for cEEG monitoring in the neuro-ICU setting have been explored. Suzuki and colleagues (68) studied patients with stenotic arterial lesions that were subjected to hypertensive and hypotensive stress while monitoring EEG frequency in the neuro-ICU. Two thirds (12 of 18) showed deterioration of the EEG after induced hypotension, and six of 11 patients showed an increase in EEG frequencies with induced hypertension. These data were considered helpful in deciding whether to place a bypass graft. Jordan and colleagues (69) described 21 stroke patients studied with cEEG and xenon-enhanced computed tomography (CT) CBF. A group with low CBF and focal slowing and a preserved background EEG improved significantly with hypervolemic therapy. Matousek and coworkers (70) correlated EEG slowing, using both CSA processing and a qualitative visual EEG assessment scale, with level of consciousness scores in comatose patients. The degree of correlation between the coma scores and EEG was highest (R = 0.68) when multiple rather than single CSA-derived EEG parameters were analyzed. Gilbert and associates (71) tested whether the EEG-derived bispectral index (BIS), an empiric index of EEG scaled from 0 to 100, correlated with neurological status in 31 awake, unsedated critically ill adults in a medical ICU. The BIS significantly correlated with a variety of neurological scores, and was more strongly correlated than other conventional CSA parameters. However, the degree of correlation was not great: In a multivariate analysis, the combination of BIS and relative theta power accounted for only 38% of the variability of Glasgow Coma Scale scores.

FIG. 8.5. Continuous electroencephalograph [EEG (cEEG)]-derived alpha-delta ratio (ADR) calculated every 15 minutes in a 57-year-old woman admitted for acute subarachnoid hemorrhage (SAH) (admission Hunt-Hess grade 4) from a right posterior communicating aneurysm (Panel A). The aneurysm was clipped on SAH day 2; no infarcts were seen on postoperative computed tomography (CT) (Fig. 8.3). She had a Glasgow Coma Scale (GCS) of 14 postoperatively. CEEG monitoring was performed from SAH day 3 to 8. The ADR progressively decreased after day 6, particularly in the right anterior region (arrow), to settle into a steady trough level later that night, reflecting loss of fast frequencies and slowing over the right hemisphere in the raw cEEG (Fig. 8.4). On SAH day 6 transcranial Doppler flow velocity in the right middle cerebral artery (MCA) was marginally elevated (144 cm/sec), but the patient remained clinically stable with hypertensive hypervolemic therapy (systolic blood pressure greater than 180 mm Hg).

Only gold members can continue reading. Log In or Register to continue