THE PROCESSED ELECTROENCEPHALOGRAM

Interpretation of the standard paper electroencephalographic tracing is a science and an art. All monitored waveforms during the case are compared with baseline signals. The interpreter has learned from experience the wide variety of normal changes that may occur in the perioperative period and promptly recognizes when changes occur that are not normal or expected. The baseline recordings and the qualitative overall impression of the record are important in interpreting the intraoperative EEG. Until recently, this qualitative approach was used because the waveforms could not be described mathematically in a timeframe that would make such information of any practical use. Analog-to-digital conversion technology associated with mainframe computers was used to convert the analog signal of the EEG to digital data and to mathematically manipulate the data. The process was complex, expensive, and still had little relevance to the clinician. Early techniques took 1 hour to digitize and analyze a 1-second epoch of electroencephalographic data. Computer hardware has dramatically improved in speed and size, and real-time signal processing is now possible and commonly used.

Several limitations are introduced when moving from the raw electroencephalographic domain to the processed electroencephalographic domain. First, as the processed electroencephalographic signal becomes more electronically remote (i.e., more processed), there is a point at which it becomes increasingly difficult or impossible to relate what we know about the raw electroencephalographic data to the processed signal. An example of this problem was found in an early prototype electroencephalographic monitor used in the operating room. When the dominant electroencephalographic frequency and amplitude were kept in an acceptable reference range, a green light was visible. When values fell outside this range, a red light appeared. Most of the valuable information contained in the EEG was not visible to the clinician, and the displayed lights were of little value in many cases. Today, some clinicians with no experience in interpreting raw electroencephalographic data are using a processed EEG with little ability to understand how it relates to the original raw data and how artifacts may contaminate the signal and appear as perfectly believable processed electroencephalographic data. Second, the standard 16-channel electroencephalographic montage provides more information than can be practically analyzed or displayed in most processed electroencephalographic monitors and perhaps more than is needed for routine intraoperative use. Studies have not elucidated the optimal number of electroencephalographic channels for intraoperative monitoring, but most available processed electroencephalographic devices use four or fewer channels of information—translating to at most two channels per hemisphere. Processed electroencephalographic devices generally monitor less cerebral territory than a standard 16-channel EEG. Third, some intraoperative changes are unilateral, and some are bilateral. Display of the activity of both hemispheres is necessary to differentiate unilateral (i.e., not caused by global factors such as anesthesia) from bilateral changes. An appropriate number of leads over both hemispheres is needed. The gold standard for intraoperative electroencephalographic monitoring is the continuous visual inspection of a 16- to 32-channel analog EEG by an experienced electroencephalographer.^{[10] [11]} Adequate studies comparing processed EEG with fewer channels to this gold standard across multiple uses and operations have not been done, although limited data using processed electroencephalographic monitoring during carotid surgery suggest that two- or four-channel instruments can detect most significant changes.^{[12] [13]}

Devices

Two basic forms of electroencephalographic processing are used: power analysis and bispectral analysis. Power analysis uses Fourier transformation to convert the digitized raw electroencephalographic signal into component sine waves of identifiable frequency and amplitude. The raw electroencephalographic data, which are plots of voltage versus time, are converted to plots of frequency and amplitude versus time. Many commercially available processed electroencephalographic machines display power (i.e., voltage or amplitude squared) as a function of frequency and time. These monitors display the data in two general forms: compressed spectral array (CSA) or density spectral array (DSA). In CSA, frequency is displayed along the x axis, and power is displayed along the y axis, with the height of the waveform equal to the

1516

Many changes that occur during anesthesia and surgery are reflected as changes in amplitude or frequency, or both. These changes can be clearly seen in the displays if adequate and appropriate channels are monitored. Power analysis has been used clinically for many years as a diagnostic tool during procedures with a risk for intraoperative cerebral ischemia such as carotid endarterectomy and cardiopulmonary bypass (CPB). Power analysis has proved to be a sensitive and reliable monitor in the hands of experienced operators using an adequate number of channels. Parameters obtained from power analysis have been investigated as monitors for depth of anesthesia. ^{[14] [15] [16] [17]} Although earlier attempts to use parameters derived from power analysis for assessment of anesthetic depth were largely unsuccessful, these same parameters are now used to various degrees with much more success as a part of different algorithms (including BIS) to measure hypnotic states.

Bispectral analysis takes into account the phase relationships between the individual components of the raw electroencephalographic signal. These phase relationships

Figure 38-4 Diagram of the technique used to generate a compressed spectral array. Below the traces, the example shows compressed spectra of the α rhythm from a normal subject. (From Stockard JJ, Bickford RG: The neurophysiology of anesthesia. In Gordon E [ed]: A Basis and Practice of Neuroanesthesia. New York, Elsevier, 1981, p 3.)