Eeg Monitoring: Intraoperative Application

ANESTHESIA FOR THE PATIENT WITH NEUROLOGIC DISEASE

0889-8537/97 $0.00

+

.20

EEG MONITORING: INTRAOPERATIVE APPLICATION Marc J. Bloom, MD, PhD

Intraoperative monitoring of electroencephalograms (EEGs) is becoming more commonplace as the technology advances: 39 making it easier and less costly to apply.48Although the cardiovascular status and systemic perfusion are routinely monitored during anesthesia, the brain, which is the target organ of general anesthesia, is generally not monitored. Instead, it is generally assumed that without prior known pathology, adequate cerebral perfusion can be assumed by maintaining adequate mean arterial pressure.72 Cerebral blood flow is not only dependent on adequate cerebral perfusion pressure, which is decreased by elevations in intracranial pressure, but also by changes in cerebral vascular resistance and cerebral vascular pathology.74Monitoring for cerebral perfusion, however, is not the only application for intraoperative EEG monitoring. EEG can also be used to monitor the depth of sedation and the degree of burst suppression produced by barbiturate infusion, as well as monitoring for the occurrence of seizures while the patients are under the influence of muscle relaxants.20Each of these applications are discussed in this article, but for intraoperative EEG to be most effectively applied, a number of fundamental principles must first be addressed. PRINCIPLES OF PROCESSED EEG

A popular misconception is that EEG is generated from a summation of action potentials propagating through axons of the brain; howFrom the Department of Anesthesiology, University of Pittsburgh Medical Center, Presbyterian University Hospital, Pittsburgh, Pennsylvania

ANESTHESIOLOGY CLINICS OF NORTH AMERICA VOLUME 15 * NUMBER 3 * SEPTEMBER 1997

551

552

BLOOM

ever, the electrical fields produced by the axonal propagation of action potentials is limited to less than 100 pm, and would be nearly impossible to measure through the scalp using surface electrodes. EEG is actually generated by the summation of potentials caused by the depolarization of the soma and apical dendrites of neurons (primarily pyramidal cells) of the outermost c0rtex.4~. 54 If the depolarization of cortical neurons was truly random and chaotic, the summation of a large number of positive and negative potentials would result in nearly complete cancellation of the measured potential at the scalp, and the resultant measured potential would be near zero. Due to tonic inhibitory influences, however, the discharge of cortical neurons is not completely chaotic, but instead is the result of the synchronous arrival of afferent depolarizations causing synaptic release of neurotransmitter in a functional subpopulation of neurons in a localized region of the cortex.” Although the depolarization of individual neurons may be as great as 100 mV, the summed potential as measured on the scalp surface is attenuated to approximately 25 to 50 FV. When inhibitory and suppressive influences, such as ischemia or drugs, are present, the frequency of neuronal depolarizations will decrease, but further synchronization may lead to initially higher amplitudes at lower frequencies in the EEG.@In conditions of cerebral excitation or disinhibition, the frequency of discharge may increase, but the desynchronization will generally result in significantly lower amplitude~.~* Diagnostic interpretation of electroencephalograms requires extensive study and training, but this is far beyond the scope of intraoperative monitoring. The intent of EEG monitoring is to detect changes in the EEG from some baseline condition, which are temporally related to particular intraopertive events or the administration of specific drugs. Although diagnostic interpretation of raw multichannel EEG involves the recognition of complex patterns, many of the salient features which characterize the acute changes in EEG can be expressed in terms of changes in amplitude and frequency content. Frequency content in the EEG has been evaluated in the past by simply counting the number of waves between fixed grid marks on the hard copy of EEG. Early efforts at automated analysis attempted to measure the “frequency of the EEG” by counting the number of times the waveform crossed zero each second, but this simplified, user-dependent interpretation produces erratic values when a combination of fast waves are superimposed on slow waves in the EEG. Now with the availability of inexpensive microcomputers, analysis of the frequency content in the EEG can easily be accomplished with the application of Fast Fourier Transformation (FFT), which converts the EEG signal from a voltage as a function of time to amplitude as a function of frequency. The FFT is a mathematical formula which closely approximates the original signal with the summation of a series of sine waves whose amplitude and phase-shift coefficients are set to produce the smallest residual error. The process is similar to taking a sine wave at each frequency and adjusting its size and shifting it along in time until it best matches the signal, and then subtracting

553

EEG MONITORING: INTRAOPERATIVE APPLICATION

that sine wave from the signal to leave a residual error which gets matched to the next higher sine wave. Alternatively, once the Fourier coefficients are known, they describe how much of the signal can be represented by a sine wave at each frequency and how much that sine wave is shifted relative to the lowest (fundamental) frequency. Squaring the complex coefficients of the Fourier transform produces the Fourier power spectrum,” a description of how much of the power in the signal is occurring at that frequency (Fig. 1). Until recently, the most advanced EEG monitoring systems simply provided cascaded spectral waveforms or topographic density plots which portrayed sequential 2to 4-second epochs of EEG activity on a video display for visual interpretation by the user.15 Even though displays of the Fourier power spectra may be easier to interpret, they still require the visual recognition of I ,

Original Signal

Digitization

A

--

One Epmh

61 Sec

1 sec

Fourier Analysis

40

1 Frequency Band IHz)

Relative Power

0.0 1.0

4.0 5.0 b.0 1.0 8.0 9.0 -

Power

2.0 3.0

f

SDectrum

D O

4

Frequency in Hz

-- 0.9 1.9

8

2.9 3.Y 4.9 5.9 6.9 1.9 8.9 9.9

Relative h p t i rude

0 0 3 6 2 3 0 2 6

0

1

Figure 1. The process of power spectrum analysis. The original continuous waveform (A) is put in digital form by sampling it repeatedly at small intervals; here 1/28th of a second. One epoch of these data (B) is then passed to a computer program to perform a Fourier analysis. The result of the analysis, shown in C, is a table giving the amplitude of the activity in each frequency band. Finally, the amplitudes are squared (to give power) and plotted as a histogram (D).(From Levy WJ, Shapiro HM, Maruchak G, et al: Automated EEG processing for intraoperative monitoring: A comparison of techniques. Anesthesiology 53:223-226, 1980; with permission.)

554

BLOOM

complex patterns by the user. Individual features in the power spectra, such as total power, median frequency, spectral edge, and relative power changes within specific banded regions of the frequency spectrum, have 7y but none of these efforts have successbeen extensively fully demonstrated a statistically reliable parameter which can be used as a criterion for the detection of specific cerebral states.', 45, 55 This is particularly unfortunate because the process of Fourier transformation itself does not reduce the massive amounts of data produced by the recording of multichannel EEG, it merely converts the description of the waveform. Furthermore, in the process of producing power spectra, all of the information about the phase relationships between frequencies is Although features such as the spectral edge frequency, which describes how high-significant frequencies can be found in the EEG, respond in general to physiologic and pharmacological changes in the brain, the EEG signal itself is the product of too many complex compo28, 42, 5s Efforts n e n t ~to~be ~ fully described by any single ~arameter.~, were also made to account for asymmetric changes between hemispheres by computation of a multiparametric asymmetry score using several features of the EEG.36Recently, however, statistical methods have been applied to extract a description of those features in the EEG76with the most discriminant value in the detection of specific cerebral changes, 83 One of the most such as ischemia* or pharmacological successful of these efforts to date has been the development of an index which uses bispectral as well as power spectral components, and describes the degree of sedative effect in the EEG.", h2 Bispectral analysis starts with a full Fourier transform and computes the degree of coherence or phase-locking between frequencies of the EEG. If a frequency component is the byproduct of the interaction of two other frequency components, the bispectrum will show a high value at the intersection of the two original frequencies. The initial version of this index was developed using the classic definition of minimum alveolar concentration (MAC) as the endpoint (i.e., "move/no move" in response to skin incision).33This index had performance problems due to the effects of opioids blunting surgical stimulation with minimal effects on the EEG.41 An improved index was developed using clinically assessed sedation as the endpoint, and this index has had remarkable success in correlating with the clinical state of the ~ a t i e n t .lo, ~ 24, , 45, h2 The performance of this index will be described later in this article. st

METHODS

When setting out to perform intraoperative EEG monitoring, one must keep in mind that attention to details can make significant differences in the effectiveness of the monitoring. When trying to generate reproducible EEG signals, it is best to place the electrodes according to the international 10/20 system, as shown in Figure 2. The international 10/20 system is based upon taking the distance from the nasion


555

Figure 2. International 10-20 system. Twenty-one electrode positions are defined by 10% or 20% of three measurements: the nasion-inion (NI) distance, the preauricular (AA') distance, and the hemicircumference (XX'). (From Kaplan JA (ed): Cardia anesthesia, ed 2. Philadelphia, WB Saunders, 1987, p 322; with permission.)

(the bridge of the nose) to the inion (the occipital base of the skull), and the distance from ear canal to ear canal over the vertex and dividing the distances along these lines at distances of lo%, 20%, and 20% to the vertex. These points are then used to produce a grid across the entire cranium. The intersection points of this grid have standardized names. By using this standardized method and its nomenclature, one is assured that electrodes can be placed in reproducible locations no matter who is doing the monitoring. Once the positions of the electrodes have been chosen, one must decide whether a referential or bipolar monitoring montage is to be applied.57Referential montages have the advantage of allowing each electrode to produce its own channel of EEG, each referenced to a common location. This reduces the number of electrodes that must be placed for multichannel recordings. Bipolar montages, on the other hand, require two electrodes for each channel, but have the advantages of better regional selectivity and lower noise, as the differential

556

BLOOM

inputs to the amplifiers reject all signals which appear in common on both electrode^.^^ The number of channels that need to be monitored is the subject of considerable debate. It is clear that early attempts at computerized EEG analysis suffered in performance from the fact that only one channel was being examined. If that channel covered only a small region of the brain, then sensitivity was lost because changes could occur in an area not being monitored. On the other hand, if the channel spanned too large an area, changes in a small region might be overshadowed by normal activity in the surrounding area. As a minimum, one must compare the left and right sides of the brain. This requires two channels as a minimum, but the regions perfused by the carotid and basilar arteries may be differently effected, and therefore the anterior and occipital regions should be monitored separately when regional differences may be important. Although four channels may be enough for adequate global sensitivity, adding more channels give better regional specificity, and most diagnostic EEG recordings are done with a minimum of 16 channels. Various types of electrodes have been used for intraoperative monitoring. Traditionally, the lowest impedance and highest signal quality have been achieved by the use of metallic cups attached to the scalp with collodion and filled with conductive gel. Although collodion and cups have the distinct advantages of low impedance and secure attachment (even in hair), the process of attachment can be time-consuming, requiring at least 1 to 2 minutes for each electrode applied. Needle electrodes can be applied much more quickly, and can be secured intraoperatively with a surgical skin staple; however, the low surface area of these electrodes tends to produce higher impedances and may be prone to polarization and DC offset when used for long periods of time. Although recent advances in the technology of conductive pads with self-adhesiveshave greatly improved their performance and simplified their use:' they cannot be applied reliably within the hair of the scalp, thereby limiting the locations available for placement. No matter what kind of electrodes are used for the monitoring, they must be applied carefully and secured extremely well to prevent intraoperative loss of satisfactory signals for monitoring. To minimize the effects of a multitude of potential artifacts and sources of noise in the operating room, the EEG signal of each channel must be preconditioned prior to digital sampling. The length of unshielded leads from the electrodes to the first amplifier should be minimized, and other electrical wires and electromagnetic sources should be kept away from the wires leading to the amplifier. The first-stage amplifiers must be completely isolated from the patient to provide electrical safety and prevent potential burns from intraoperative electrocautery, and the amplifiers should be capable of handling extensive saturation, which may be caused by high-amplitude signals from sources such as electrocautery.80Relatively narrow frequency filtering is often necessary to produce satisfactory EEG recordings in the operating room. Low-


557

frequency cut-offs typically range from 0.3 to 2 Hz, and high-frequency cut-offs typically range from 30 to 50 Hz. If contamination from AC power is problematic, a 60-Hz notch filter can also be added. Ideally, all these filters should have minimum or zero phase shift, which is most easily achieved by digital filtering techniques after analog to digital con~ersion.~~ When selecting a set of locations for electrodes (a montage), the principle to keep in mind is that the electrodes should straddle the area of the brain one wishes to monitor for a particular physiologic change. This is of less concern for global processes, such as global ischemia or the effects of sedative drugs, but is particularly important in preserving the sensitivity and specificity of monitoring for localized cerebral ischemia. A study by Craft et all6 found that 4-channel EEG monitoring was 100% sensitive in detection of ischemia from carotid cross-clamping when all of the electrodes were kept within the perfusion area of the middle cerebral artery, but the sensitivity fell off slightly when the electrodes in the montage were placed in areas perfused by other arteries. When configuring an automated EEG monitor, amplification and display ranges should be kept as large as possible without significant saturation of the signal to provide maximum sensitivity, and update rates should be kept relatively rapid in comparison to the time course of the expected EEG changes. That is to say, for ischemia monitoring updates should be made more frequently then every 10 seconds, while when monitoring for pharmacological actions, updates less frequent than 1 per minute may be satisfactory. When employing any form of processed EEG monitoring, care must be taken that the EEG signal is of high quality prior to further levels of processing and abstraction, otherwise the risk is that a noisy signal contaminated with significant artifacts may produce a display subject to significant misinterpretation.' Common sources of noise are shown in the Figure 3. AC power noise is a common contaminant, which can be minimized by good isolation, the use of well-shielded power supplies, keeping power cords well-removed from the unamplified EEG signal, the use of differential inputs to the amplifier, and the use of an AC notch filter when necessary. EKG contamination can best be minimized by avoiding electrodes placed low on the head or neck, minimizing electrode impedances, use of differential inputs to bipolar montages, and in some cases, by repositioning electrodes of problematic channels. Motion and microphonic artifacts can best be reduced by minimizing lead wire contact with repetitively moving objects, such as ventilation hoses and other sources of vibration. Electrode-related noises and other artifacts can be minimized by using electrodes with large surface areas, thoroughly cleaning reusable electrodes, using only one kind of an electrode in a montage, and minimizing impedances of the electrodes. Roller pumps, particularly those used in cardiac surgery, can produce particularly problematic vibration at frequencies within the range of interest for intraoperative EEG.40These artifacts may be particularly

558

BLOOM

Artifact

Voltage Waveform

TYF

Frequency Spectrum

ECG

Swdlow 0:

Blink

“Sovie”

Tr; bo-

A

”

Elccric

k-4 1 second

Figure 3. Stylized artifacts are illustrated in both time and frequency domains. Electromyographic artifact generally adds uniform wideband (white) noise. Cardiac electrical activity adds a Q R S pulse in the time domain that translates to a series of blips in the frequency domain spaced at intervals of the heart rate frequency. Swallows, blinks, and motion artifact are heavily weighted toward delta frequencies. Electrosurgical artifact (Bovie) when present in small amounts is white noise; if stronger, it saturates the EEG amplifier producing flat line EEG tracings. Power line interference may significantly distort the time domain waveform, but generally leaves the spectral display unaltered, save a blip at 60 Hz. The triboelectric effect is created by mechanically deforming an insulator around a conductor, e.g., a cardiac bypass roller pump creates the rhythmic pattern in the figure, which is then conducted by the blood into the body. (From Rampil IF: What every neuroanesthesiologist should know about electroencephalograms and computerized monitors. In Bissonnette B (ed): Cerebral Protection, Resuscitation, and Monitoring: A Look in the Future of Neuroanesthesia. Philadelphia, WB Saunders, 1996, pp 683-718; with permission.)


559

difficult to eliminate, but one can try moving the pumps as far from the recording areas as possible or noting the characteristic frequencies produced when the pumps are switched on. Efforts to automatically eliminate common artifacts have met with some success, and continue to be d e ~ e l o p e d . ~ The choice of the display formats and settings is often a matter of personal preference, but most of the intraoperative EEG monitoring equipment produces similar choices of compressed spectral array or 39 from relatively short density spectral array displays of power ~pectra,"~, 2- to 4-second epochs of raw EEG signals.38Update rates of the displays should be adjusted so that response times are satisfactory while providing several minutes of comparative data to detect the onset of changes. It is preferable that representative portions of the raw waveform are continuously presented on the monitor to assure that a high-quality signal is being analyzed without significant artifact. The derivation of a discriminant function to describe EEG changes takes automated processing a step beyond simply transforming the data. It applies statistical techniques to classify a set of parameters according to predefined conditions. In the case of the Bispectral Index (BIS), bispectral components,46which measure the degree of phase-locking between frequencies in the EEG, are combined with power spectral components into a f~nction,6~ which is highly correlated with the degree of sedation produced by drugs (Fig 4).h3,81 This process represents a distinct difference from prior methods which chose a prioui, a parameter thought to be useful, and then tested its performance. Such methods showed qualitative suggestion of a particular effect on the EEG, but failed to stand

Correlate Best with Clinical Endpoints

I

I

/Combine Features to Produce1 Bispectrai index Figure 4. Measuring the effect of anesthetics on the brain, technology development process. The BIS index is created by discriminant analysis of features from both the bispectrum and the power spectrum. The index is further enhanced by specific EMG and ECG artifact elimination, and burst suppression detection. The result is an index with very high correlation to clinically assessed sedation levels.

560

BLOOM

up to rigorous statistical testing. Building a discriminant function designed to have high correlation with the observed clinical condition has given the anesthesiologist a ”window on the brain” to quantify the EEG effects of sedative drugs. APPLICATIONS OF PERIOPERATIVE EEG MONITORlNG Depth of Sedation

While the site of action of anesthetic drugs is the brain, the individual requirements for adequate anesthetic drug dosing has until recently only been assessed by indirect measures of the effects on the cardiovascular system and other autonomic responses. Because individual requirements may vary by as much as 200% of the population averages, it is only the relatively large therapeutic index of most anesthetic drugs that makes it possible to administer anesthetic agents based solely on estimates of the requirements from population statistic^.'^ While qualitative correlations of simple parameters of EEG, such as median frequency 27 have been demon~trated,~~ the high variance and and spectral low specificity of these measures have made them inadequate as a reliable means of assessing an individual patient’s requirement^.'^ Recently, however, efforts to statistically derive discriminant functions from the EEG which correlate with specific cerebral states76have met with significant success.62In particular, a discriminant function has been developed which includes not only power spectral components, but parameters from bispectral analysis as well.64This provides additional information above and beyond that of the power spectrum, which ignores all phase information. The BIS has been specifically designed to reflect the sedative effect of a wide range of drugs across the complex and multiphasic response ranges by including various components, such as beta activation at light levels of sedation and burst suppression at the deepest levels of sedation, all in one discriminant function (Fig. 5). This function has shown high correlation with clinically assessed levels of ~ e d a t i o n Monitoring .~~ the bispectral index as a measure of the depth of sedation has clinical utility throughout the period of perioperative management. BIS monitoring can objectively assess the adequacy of sedation prior to induction of anesthesia and aid in the assessment of anesthetic requirements during induction by detecting patients either overly sensitive or tolerant to anesthetic agents. BIS also demonstrates the significant differences in the time of onset of these drugs in various patients. Because the inadequacy of intraoperative sedation is generally difficult to assess, particularly when muscular blocking agents are used, clinicians generally administer a relative overdose of anesthetic agents to most patients unless limited by untoward side effects, such as hypotension. It has been found that by directly measuring the brain’s response to sedative agents one can safely and effectively reduce the


0

20

40

60

80

561

100

Bispectral Index 3.0 Figure 5. Actual logistic regression curves of the probability of consciousness from the data of volunteers anesthetized with standardized protocols. Solid curve = isoflurane; shaded curve = midazolam; and circled curve = propofol. The probability of consciousness falls at a slightly lower BIS for propofol, but this was not statistically significant.

dosage of sedative agents in the vast majority of patients. The time of emergence and the required periods of postoperative recovery are both shortened while suffering no increase in undesired clinical response^.^'" This is particularly useful in those patient populations where rapid emergence is of significant benefit, such as in neurosurgical patients that need to have their neurologic status assessed as soon as possible, and in the ambulatory surgical setting where one wishes to discharge patients as quickly as possible. Even in those settings where full general anesthesia is not applied, such as in regional anesthetics and in cases involving monitoring with sedation, it is useful to have an objective measure of the degree of sedation, thereby minimizing the potential for cumulative effects to produce relative overdoses and the consequences of Although the incidence of intraoperative awareness is extremely low, it continues to occur with disturbing regularity. Although its effectiveness has not been demonstrated, one would expect that monitoring of the BIS would provide a significant additional assurance that patients are receiving sufficient doses of sedative to prevent intraoperative awareness, even during techniques which employ hgh-dose narcotics and muscle relaxants which have the highest incidences of such occurrence^.^^ Monitoring of sedation levels by use of the BIS also has utility in the intensive care unit, particularly when patients are under the influence of neuromuscular relaxants or those under very heavy Using

562

BLOOM

BIS monitoring in these patients avoids the problem of having patients generally overdosed because the level of their sedation cannot be determined.8zMaintaining a light but adequate dose of sedation allows for the rapid withdrawal of sedation in order to provide an opportunity for clinical assessment as well as avoiding any problems associated with long-term drug accumulation. One further application of bispectral measurement of the level of sedation is for investigation into the pharmacodynamics of sedative drugs by providing a quantitative measure of the central nervous system response to known dosages. Each patient’s pharmacodynamics can be determined. This allows the design of individualized dosing schemes rather than those based on population averages (Fig. 6). To properly use the BIS to dose sedative drugs, one must understand what BIS is not. BIS is not a measure of MAC (i.e., it does not predict whether a patient will react to a surgical stimulus). It is also not specifically designed to detect ischemia, but may decrease if ischemia causes a change in conciousness. Lastly, like any other monitor, it is not a substitute for sound clinical judgement. Assessment of Cerebral Well-Being and the Detection of Cerebral Ischemia

It has long been appreciated that cerebral ischemia is reflected in changes in the electroen~ephalogram’~~ 77, 74, ”; however, although ischemia can be reliably detected by an experienced electrophysiologist, a reliable single quantitative descriptor of cerebral ischemia has yet to be described.55,7M Nonetheless, power spectral displays of EEG have been shown to reliably detect cerebral ischemia when the electrodes are

Analgesics/ Local Blocks (Sensory Attenuation)

Anesthetics/ Sedatives (Depth of Hypnosis)

Figure 6. Guidelines for the clinical application of BIS, indicating the need for a balance between analgesia and sedation. Both factors must be addressed in anesthetic management. The balance between stimulation and sensory suppression is critical. Patient arousal can occur at any hypnotic level if stimulation is not blocked. Patients can be maintained at light hypnotic levels (BIS, 55-65) if the perceived levels of stimulation are minimal. Changes in BIS in response to stimulation can provide an alternative real-time measure of patient reactiveness under anesthesia.

EEG MONITORING: INTRAOPERATIVE API'LICATlON

563

placed entirely within the area subject to ischemia. In those situations where cerebral ischemia may represent a significant risk to the perioperative patient, there appears to be good reason to apply EEG monitoring to provide an early warning of the onset of cerebral ischemia, thereby allowing efforts to improve cerebral perfusion and oxygen delivery before irreversible damage occurs (Fig. 7). Smoothing Rates: Bispectral: 60 Spectral: 30

Filters: High: 50 Hz Notch: 60 HZ Law: 0.25 HZ 1 1

Ch3TP3-C31

Edge 5EF Montage: Bipolar

L AMPLITUDE

R

16

Ch4fP4-C41 0 -10 20

31

Figure 7. DSA display of 4-channel EEG during placement of an Automatic Implantable Cardiac Defibrillator (AICD). Two short periods of systemic hypotension are seen at 8:lO and 8:12 following induction. A brief ischemic episode is also seen at 8:44 when ventricular fibrillation is induced to test the device.

564

BLOOM

The most obvious application of monitoring for cerebral ischemia is in the case of carotid endartere~tomy.~~ Although the insertion of an intravascular shunt can provide alternative flow when inadequate perfusion occurs during carotid cross-clamping, insertion of a shunt carries its own risks of embolization and m a l f ~ n c t i o nThe . ~ ~ shunt also limits exposure of the artery, making an optimal intimal repair more difficult. By detecting the 10% to 15% of patients who do not tolerate carotid cross-clamping without shunting, 85% of carotid endarterectomy patients are spared the added risk of shunt insertion.58Several studies have now found that selective shunting based on EEG monitoring can significantly reduce cerebral morbidity associated with carotid endarterectomy.l5r21, 61 EEG changes associated with carotid cross-clamping are usually not subtle, and can be readily seen on most spectral displays (Fig. 8). The temporal association of any changes persisting for more than 20 seconds after the cross-clamp in the absence of other physiologic or pharmacologic changes can be interpreted as indicative of ischemia, and may warrant shunting. Furthermore, because the vast majority of postoperative complications after endarterectomy are associated with hypertension, EEG allows for the aggressive but rational control of blood pressure, thereby minimizing the risks of postoperative hypertension. Quantitative indices similar to those developed for the bispectral index of sedation are currently in development for the detection of cerebral ischemia. Another area where EEG monitoring for ischemia appears to hold /-I. Chl :F3-C3

LEFT

15 :26 15:27 15 :28 15 :29 15 :38 15:31 30

Chl :F3-C3

-2.

20

10

0

5HpU/d iu 25

0

Ch2 :F4-C4 mm/s

20

38

‘3.

Figure 8. DSA display during cerebral ischemia induced by cross-clamping of the right carotid artery. (1) Decreased density in the power spectra on the right side; (2) decreased amplitude in the right amplitude bar; and (3) decreased voltage in the raw EEG waveform on the right side. Note that since all frequencies have decreased proportionately in amplitude, the Spectral Edge Frequency (SEF) remains constant.


565

promise is during cardiac surgery.s6,59, 6o Various studies have found neuropsychiatric deficits in as high as 40% to 70% of patients following cardiac s ~ r g e r y , 6 ~ but ~ ' a study by Edmonds et all8 has found that neuropsychiatric deficits in their population of patients could be lowered from nearly 30% to less than 4% by aggressive intervention when intraoperative changes in EEG suggested cerebral ischemia.2,18, 19, 2s Some studies' have pointed out that automated EEG analysis must be applied with caution because although sensitivity has been found to be very high, specificity can be severely affected by confounding variables, such as anesthetic technique and analysis methods. Automated artifact detection and rejection are also limited in their effectiveness, and can lead to misinterpretation if taken entirely at face value. One must realize that EEG changes associated with cerebral ischemia occur at variable times and to a variable degree. If the precautions outlined in the methods section are taken and quantitative measures are interpreted only as another of the imperfect parameters monitored during anesthesia, EEG monitoring can be expected to decrease CNS morbidity by detecting potential conditions of cerebral ischemia. When monitoring for cerebral ischemia, interventions based solely on current quantitative EEG parameters remain imprudent at this time. Cerebral Vascular Surgery

Cerebral vascular surgery, such as that for cerebral aneurysms and arteriovenous malformations, may produce distinct and identifiable periods during which cerebral perfusion may be compromised. EEG monitoring provides a way of detecting and assessing the degree of ischemia associated with such maneuvers as retraction and temporary clipping of proximal vessels. EEG monitoring may help to assure the patency of critical vessels after aneurysm clipping or demonstrate adequate collateral perfusion when cerebral vessels must be ~acrificed.~" One final potential use of EEG is in those situations where chronically hypertensive patients are exposed either deliberately or unavoidably to relative hypotension.60It is well known that chronically hypertensive patients tend to shift their autoregulatory curve to within the range of blood pressures they are usually exposed to. Because the lower end of their autoregulatory mechanism may have moved upward, these patients may be at risk for relative cerebral hypoperfusion at blood pressures that would be satisfactory to the normotensive population. EEG monitoring may provide an added measure of safety in management of these patients by alerting the anesthesiologist to a condition that might not otherwise be appreciated." Assessing Efforts of Cerebral Protection and Barbiturate Coma

While the actual efficacy of barbiturate coma for cerebral protection is still being debated, the practice of induction of barbiturate coma is

566

BLOOM

relatively common.47Studies have found that the necessary dose of barbiturate to induce burst suppression varies greatly from individual to individual in a range from as little as 2 mg/kg to as much as 25 mg/ kg. With such a wide range of dosage necessary for the desired effect, it is clear that the only rational way to administer barbiturates to the level of burst suppression is to measure the degree of burst suppression using EEG (Fig. 9).,, Many of the processed EEC systems now available provide quantitation of the percentage of time that the EEG is suppressed, thereby providing direct and immediate feedback on the effects of barbiturate infusion and minimizing the amount of drug that must be infused while insuring the achievement of the desired result. Detection of Seizures

Although seizures are a relatively uncommon intraoperative event, the consequences can be devastating, and without EEG, they may be impossible to detect, particularly when neuromuscular relaxants are employed. In those patients with known or potential risks of intraoperative seizures,14it would be prudent to apply EEG monitoring, as seizure activity is evident on EEG, even to relatively inexperienced users.5o

1.

2.

C

C

r

5EpUld i u

Figure 9. DSA display during coronary artery bypass surgery showing burst suppression pattern induced for cerebral protection. (1) DSA plot showing bursts of activity across a wide range of frequency altering with periods of little or of no EEG activity; (2) amplitude bar with alternating high and low amplitudes; and (3) raw EEG waveform with bursts of EEG activity alternating with little or no apparent EEG activity.

EEG MONITORING: INTRAOPERATIVE APPLICATlON

567

Conclusion As the technology advances, EEG monitoring gets easier to use and more affordable. With the understanding of a few basic principles and attention to the details of good recording techniques, perioperative EEG monitoring has several clinical applications. Statistical analyses, such as the BIS index, make measurement of the level of sedation a useful and practical tool in the dosing of anesthetic drugs. EEG helps to detect acute cerebral ischemia and the need for shunting in cerebrovascular surgery. EEG monitoring is necessary for the proper application of barbiturate coma. EEG may help to decrease the incidence of neuropsychiatric disorders after cardiac surgery. Although the utility of EEG monitoring has been debated in the past, recent advances make it likely that perioperative use of EEG will become more common.

References 1. Adams DC, Heyer EJ, Emerson RG, et al: The reliability of quantitative electroencephalography as an indicator of cerebral ischemia [see comments]. Anesth Analg 81:SO83, 1995 2. Arom KV, Cohen DE, Strobl FT: Effect of intraoperative intervention on neurological outcome based on electroencephalographic monitoring during cardiopulmonary bypass. Ann Thorac Surg 48:476483, 1989 3. Barlow JS, Creutzfeldt OD, Michael D, et al: Automatic adaptive segmentation of clinical EEGs. Electroencephalogr Clin Neurophysiol 51:512-525, 1981 4. Barlow J S Artifact processing (rejection and minimization) in EEG data processing. In Lopes da Silva FH, Storm van Leeuwen W, Remond A (eds): Clinical Applications of Computer Analysis of EEG and other Neurophysiological Signals. Amsterdam, Elsevier, 1986, pp 15-62 5. Bashein G, Nessly ML, Bledsoe SW, et al: Electroencephalography during surgery with cardiopulmonary bypass and hypothermia [see comments]. Anesthesiology 76~878-891, 1992 6. Bender R, Schultz B, Schultz A, et al: Identification of EEG patterns occurring in anesthesia by means of autoregressive parameters. Biomedizinische Technik 36236240, 1991 7. Bickford RG, Melby G, Johnson S, et al: Computer automated EEG, diagnostic and interpretation system (CAEDIS): Pilot studies with a hybrid system. Electroencephalogr Clin Neurophysiol 30:95, 1971 8. Bloom MJ: Techniques to identify clinical contexts during automated data analysis. Int J Clin Monit Comput 10:17-22, 1993 9. Bloom MJ, Whitehurst S, Mandel M, et al: Bispectral Index as an EEG measure of the sedative effect of isoflurane [abstract]. Anesthesiology 83:A195, 1995 10. Bloom MJ, Whitehurst S, Mandel M, et al: Bispectral Index as an EEG measure of sedation with methohexital [abstract]. Anesthesiology 85:A461, 1996 11. Bodenstein G, Schneider W, Malsburg CV: Computerized EEG pattern classification by adaptive segmentation and probability-density-function classification: Description of the method. Comput Biol Med 15297-313, 1985 12. Boeijinga PH, Lopes da Silva FH: Differential distribution of beta and theta EEG activity in the entorhinal cortex of the cat. Brain Res 448:272-286, 1988 13. Brundidge PK, Leave11 ME, Tempelhoff R EEG-controlled "overdosage" of anesthetics in a patient with a history of intra-anesthetic awareness [see comments]. J Clin Anesth 6:496-499. 1994

568

BLOOM

14. Cascino GD, So EL, Sharbrough FW, et al: Alfentanil-induced epileptiform activity in patients with partial epilepsy. J Clin Neurophysiol 10:520-525, 1993 15. Chiappa KH, Burke SR, Young RR Results of electroencephalographic monitoring during 367 carotid endarterectomies: Use of a dedicated minicomputer. Stroke 10:381388, 1979 16. Craft RM, Losasso TJ, Perkins WJ, et al: EEG monitoring for cerebral ischemia during carotid endarterectomy (CEA): How much is enough [abstract]? Anesthesiology 81:A214, 1994 17. Dwyer RC, Rampil IJ, Eger EI, et al: The electroencephalogram does not predict depth of isoflurane anesthesia. Anesthesiology 81:403409, 1994 18. Edmonds HL Jr, Griffiths LK, van der Laken J, et al: Quantitative electroencephalographic monitoring during myocardial revascularization predicts postoperative disorientation and improves outcome [see comments]. J Thorac Cardiovasc Surg 103:555563, 1992 19. Edmonds HL Jr, Rodriguez RA, Audenaert SM, et al: The role of neuromonitoring in cardiovascular surgery [review]. J Cardiothorac Vasc Anesth 1015-23, 1996 20. Emerson RG, Turner CA: Monitoring during supratentorial surgery [review]. J Clin Neurophysiol 10:404411, 1993 21. Facco E, Deriu GP, Dona B, et al: EEG monitoring of carotid endarterectomy with routine patch-graft angioplasty: An experience in a large series. Neurophysiol Clin 22:437446, 1992 22. Frawley JE, Hicks RG, Horton DA, et al: Thiopental sodium cerebral protection during carotid endarterectomy: Perioperative disease and death. J Vasc Surg 19:732-738,1994 23. Gaitini L, Vaida S, Collins G, et al: Awareness detection during caesarean section under general anaesthesia using EEG spectrum analysis. Can J Anaesth 42377-381, 1995 24. Glass P, Can TJ, Sebel PS, et al: Comparison of the Bispectral Index (BIS) and measured drug concentrations for the monitoring of effects of propofol, midazolam, alfentanil, and isoflurane [abstract]. Anesthesiology 83:A374, 1996 25. Grote CL, Shanahan PT, Salmon P, et al: Cognitive outcome after cardiac operations: Relationship to intraoperative computerized electroencephalographic data. J Thorac Cardiovasc Surg 1041405-1409, 1992 26. Gurman GM: Assessment of depth of general anesthesia: Observations on processed EEG and spectral edge frequency. Int J Clin Monit Comput 11:185-189, 1994 27. Gurman GM, Fajer S, Porat A, et al: Use of EEG spectral edge as index of equipotency in a comparison of propofol and isoflurane for maintenance of general anaesthesia. Eur Anaesthesiol 11:443-448, 1994 28. Hanowell LH, Soriano S, Bennett HL: EEG power changes are more sensitive than spectral edge frequency variation for detection of cerebral ischemia during carotid artery surgery: A prospective assessment of processed EEG monitoring. J Cardiothorac Vasc Anesth 6292-294, 1992 29. Jansen C, Moll FL, Vermeulen FE, et al: Continuous transcranial Doppler ultrasonography and electroencephalography during carotid endarterectomy: A multimodal monitoring system to detect intraoperative ischemia. Ann Vasc Surg 795-101, 1993 30. Jones TH, Chiappa KH, Young RR, et al: EEG monitoring for induced hypotension for surgery of intracranial aneurysms. Stroke 10:292-294, 1979 31. Jopling M, Lang E, Embree P, et al: An evaluation of the aspect "ZIPprep" surface electrode [abstract]. Anesth Analg 80:SCA90, 1995 32. Kearse LA Jr, Lopez-Bresnahan M, McPeck K, et al: Preoperative cerebrovascular symptoms and electroencephalographic abnormalities do not predict cerebral ischemia during carotid endarterectomy. Stroke 261210-1214, 1995 33. Kearse LA Jr, Manberg P, Chamoun N, et al: Bispectral analysis of the electroencephalogram correlates with pateint movement to skin incision during propofol/nitrous oxide anesthesia. Anesthesiology 81:1365-1370, 1994 34. Kearse LA Jr, Marota :'IJ The role of the electroencephalogram in assessing memory function under general anesthesia [editorial; comment]. J Clin Anesth 6:457461, 1994 35. Kearse LA Jr, Martin D, McPeck K, et al: Computer-derived density spectral array in


569

detection of mild analog electroencephalographic ischemic pattern changes during carotid endarterectomy. J Neurosurg 78:884-890, 1993 36. Kopruner V, Pfurtscheller G: Multiparametric asymmetry score (MAS): Distinction between normal and ischaemic brains. Electroencephalogr Clin Neurophysiol57343346, 1984 37. Levy WJ: Intraoperative EEG patterns: Implications for EEG monitoring. Anesthesiology 60:430434, 1984 38. Levy WJ: Effect of epoch length on power spectrum analysis of the EEG. Anesthesiology 66:489495,1987 39. Levy WJ, Shapiro HM, Maruchak G, et al: Automated EEG processing for intraoperative monitoring: A comparison of techniques. Anesthesiology 53:223-236, 1980 40. Levy WJ, Shapiro HM, Meathe E: The identification of rhythmic EEG artifacts by power-spectrum analysis. Anesthesiology 53:505-507, 1980 41. Liu PW, Wang H, Kearse LA Jr: The Bispectral Index does not predict hemodynamic responses to laryngoscopy and intubation during propofol-induced EEG burst suppression [abstract]. Anesthesiology 85:A175, 1996 42. Lopes da Silva FH: Analysis of EEG non-stationarities [review]. Electroencephalogr Clin Neurophysiol (supp1)163-179, 1978 43. Lopes da Silva F Neural mechanisms underlying brain waves: From neural membranes to networks [review]. Electroencephalogr Clin Neurophysiol 79%-93, 1991 44. Lopes da Silva F, Pijn JP, Boeijinga P: Interdependence of EEG signals: Linear vs. nonlinear associations and the significance of time delays and phase shifts. Brain Topogr 2:9-18, 1989 45. Matthis P, Scheffner D, Benninger C: Spectral analysis of the EEG: Comparison of various spectral parameters. Electroencephalogr Clin Neurophysiol 52:218-221, 1981 46. Muthuswamy J, Roy R, Sharma A: A study of electroencephalographic descriptors and end-tidal concentration in estimating depth of anesthesia. J Clin Monit 12:353364, 1996 47. Nussmeier NA, Arlund C, Slogoff S: Neuropsychiatric complications after cardiopulmonary bypass: Cerebral protection by a barbiturate. Anesthesiology 64:165-170, 1986 48. Nuwer MR: Intraoperative electroencephalography [review]. J Clin Neurophysiol 10:437444, 1993 49. Nuwer MR Electroencephalograms and evoked potentials: Monitoring cerebral function in the neurosurgical intensive care unit [review]. Neurosurg Clin North Am 5~647-659,1994 50. Nuwer MR Quantitative EEG analysis in clinical settings. Brain Topogr 8:201-208, 1996 51. Nuwer MR, Daube J, Fischer C, et al: Neuromonitoring during surgery: Report of an IFCN Committee [see comments]. Electroencephalogr Clin Neurophysiol 87263-276, 1993 51a. Payne FB, Sebel MB, Glass PSA, et al: Bispectral Index (BE) monitoring allows faster emergence from propofol alfentanil/N20 anesthesia. Anesthesiology 85:A1056, 1996 52. Pfurtscheller G, Aranibar A: Event-related cortical desynchronization detected by power measurements of scalp EEG. Electroencephalogr Clin Neurophysiol 42:817826, 1977 53. Pivik RT, Broughton RJ, Coppola R, et al: Guidelines for the recording and quantitative analysis of electroencephalographic activity in research contexts [review]. Psychophysiology 30:547-558, 1993 54. Prior PF: EEG monitoring and evoked potentials in brain ischaemia [review]. Br J Anaesthesia 5763-81, 1985 55. Pronk RA, Simons AJ: Processing of the electroencephalogram in cardiac surgery. Computer Programs in Biomedicine 18:181-189, 1984 56. Pronk RA, Simons AJ, De Boer SJ: The use of the EEG for patient monitoring during open heart surgery. pp. 77-81. In: Anonymous Calif., IEEE Computer Society, 1975. WG 141 C736 1975. In: Computers in Cardiology. Long Beach, 1996 57. Rampil IJ: What every neuroanesthesiologist should know about electroencephalograms and computerized monitors. In Bissonnette B (ed): Cerebral Protection, Resusci-

570

58. 59. 60. 61. 62. 63. 64. 65. 66. 67. 68. 69. 70. 71. 72. 73. 74. 75. 76. 77. 78. 79. 80.

81.

BLOOM

tation, and Monitoring: A Look in the Future of Neuroanesthesia. Philadelphia, WB Saunders, 1996, pp 683-718 Redekop G, Ferguson G: Correlation of contralateral stenosis and intraoperative electroencephalogram change with risk of stroke during carotid endarterectomy [see comments]. Neurosurgery 30:191-194, 1992 Salerno TA, Lince DP, White DN, et al: Monitoring of electroencephalogram during open-heart surgery: A prospective analysis of 118 cases. J Thorac Cardiovasc Surg 76:97-100, 1978 Salemo TA, Lynn RB, White DN, et a1 Cerebral surveillance during cardiac surgery. Can J Surg 22325-326, 1979 Sbarigia E, Speziale F, Colonna M, et a1 The selection for shunting in patients with severe bilateral carotid lesions. Eur J Vasc Surg 7(suppl A):3-7, 1993 Sebel PS, Bowles SM, Saini V, et al: EEG bispectrum predicts movement during thiopental/isoflurane anesthesia. J Clin Monit 11333-91, 1995 Sebel PS, Bowles SM, Saini V, et al: EEG Bispectrum predicts movement during thiopental/isoflurane anesthesia. J Clin Monit 11:83-91, 1995 Sigl JC, Chamoun NG: An introduction to bispectral analysis for the electroencephalogram. J Clin Monit 10:392404, 1994 Sotaniemi K: Prediction of cerebral outcome after extracorporeal circulation. Acta Neurol Scand 66:697-704, 1982 Sotaniemi K: Five-year neurological and EEG outcome after open-heart surgery. J Neurol Neurosurg Psychiatry 48:569-575, 1985 Sotaniemi KA: Brain damage and neurological outcome after open-heart surgery. J Neurol Neurosurg Psychiatry 43:127-135, 1980 Sotaniemi KA: The benefits of open-heart surgery as reflected in the EEG. Scand J Thorac Cardiovasc Surg 15:205-212, 1981 Sotaniemi KA: Cerebral outcome after extracorporeal circulation: Comparison between prospective and retrospective evaluations. Arch Neurol 40:75-77, 1983 Sotaniemi KA, Juolasmaa A, Hokkanen ET: Neuropsychologic outcome after openheart surgery. Arch Neurol 385-8, 1981 Sotaniemi KA, Sulg IA, Hokkanen TE: Quantitative EEG as a measure of cerebral dysfunction before and after open-heart surgery. Electroencephalogr Clin Neurophysiol 50:81-95, 1980 Stockard JJ, Bickford RG, Schauble JF: Pressure-dependent cerebral ischemia during cardiopulmonary bypass. Neurology 23:521-529, 1973 String ST, Callahan A: The critical manipulable variables of hemispheric low flow during carotid surgery. Surgery 93:4649, 1983 Sundt TM, Sharbrough FW, Piepgras DG: Correlation of cerebral blood flow and electroencephalographic changes during carotid endarterectomy with results of surgery and hemodynamics of cerebral ischemia. Mayo Clin Proc 56:533-543, 1981 Thomas WA, Cole PV, Etherington NJ, et a1 Electrical activity of the cerebral cortex during induced hypotension in man: A comparison of sodium nitroprusside and trimetaphan. Br J Anaesth 57:134-141, 1985 Thomsen CE, Rosenfalck A, Norregaard Christensen K: Assessment of anaesthetic depth by clustering analysis and autoregressive modelling of electroencephalograms. Comput Methods Programs Biomed 34:125-138, 1991 Tolonen U: Quantitative 99mtechnetium cerebral circulation time in brain infarction: Its relation to clinical findings, electroencephalograms and conventional radionuclide studies. Acta Neurol Scand 64~337-352, 1981 Tolonen U: Parametric relationships between four different quantitative EEG methods in cerebral infarction. Prog Brain Res 6251-64, 1984 Tolonen U, Sulg IA: Comparison of quantitative EEG parameters from four different analysis techniques in evaluation of relationships between EEG and CBF in brain infarction. Electroencephalogr Clin Neurophysiol 51:177-185, 1981 van der Weide H, Pronk RA: Interference suppression for EEG recording during open heart surgery. Electroencephalogr Clin Neurophysiol46:609-612, 1979 Vernon JM, Lang E, Sebel PS, et al: Prediction of movement using bispectral EEG


571

analysis during propofol/alfentanil of isoflurane/alfentanil anesthesia. Anesth Analg 801780-785, 1995 82. Veselis RA, Reinsel R, Marino P, et al: The effects of midazolam on the EEG during sedation of critically ill patients. Anaesthesia 48:463470, 1993 83. Veselis RA, Reinsel R, Sommer S, et al: Use of neural network analysis to classify electroencephalographic patterns against depth of midazolam sedation in intensive care unit patients. J Clin Monit 7:259-267, 1991 84. Veselis RA, Reinsel RA, Wronski M, et al: EEG and memory effects of low-dose infusions of propofol. Br J Anaesth 69:246-254, 1992

Address reprint requests to Marc J. Bloom, MD, PhD Department of Anesthesiology University of Pittsburgh Medical Center C200-Presbyterian University Hospital 200 Lothrop Street Pittsburgh, PA 15213

Eeg Monitoring: Intraoperative Application

Recommend Documents