METHODS 25, 272–287 (2001) doi:10.1006/ meth.2001.1239, available online at http:/ / www.idealibrary.com on Real-Time Intraoperative Neurophysiological Monitoring Don Krieger and Robert J. Sclabassi Department of Neurosurgery, University of Pittsburgh, 200 Lothrop Street, Pittsburgh, Pennsylvania 15213 must be enhanced. The resultant waveforms must then We discuss the utilization of signal processing techniques during be displayed and stored. For evoked responses, appro- surgical procedures. These techniques are used to provide real- priate stimuli must be generated and applied to excite time monitoring of nervous system function. We describe the the nervous system with accurate synchronization with historical development of these techniques and the hardware and waveform sampling. This sequence is presented graphi- software that have been used to implement them.  2001 cally in Fig. 2. Signal conditioning may occur both be- Elsevier Science fore and after amplification. Note too that signal/noise enhancement is not always required. Transmission is required when the signal display/storage is located at a distance from the apparatus responsible for the earlier The human nervous system produces a variety of steps. Signal storage is often a requirement due to the minute but measurable electrical signals. These sig- need for detailed records both for training and for nals, when interpretable, may be used to continuously medical/legal documentation. monitor nervous system function. Short-latency feed- This entire sequence must occur in “real time.” That back may then be provided to the surgeon to help guide is, while there will be a finite lag between signal acquisi- his or her manipulation of neural tissue. This feedback tion and interpretation, this lag must remain approxi- loop can be critical in detecting and avoiding or re- mately constant. Because human interaction is re- versing both surgically induced and spontaneous injury. quired for signal interpretation, the shortest latency An example of the former is infarction of either brain for the entire sequence is approximately 1 s. It may, or spinal cord due to retraction. An example of the latter however, be up to a minute or even longer depending is infarction of the brain due to spontaneous propaga- on the quantity of data required to produce adequate tion of an atherosclerotic plaque from the carotid artery. signal/noise enhancement. Typically the lag is 5–60 s Consider the structure of the nervous system first for brain activity. (Fig. 1) and the related inventory of signals and their In addition to evoked potientals, there are many pro- properties (Table 1). For those procedures for which cedures in which it is of significant clinical value to monitoring of either spontaneous or evoked activity is monitor the ongoing electroencephalogram (EEG) and useful, it is the evoked responses that typically provide the spontaneous electromyogram (EMG). The EEG/ the most consistent, specific, readily interpretable, and EMG is the spontaneous electrical activity recordable most sensitive information. For the spinal cord, periph- from the brain/muscle. Arithmetically derived versions eral nerves, muscles, and sensory receptors, the evoked of the EEG, e.g., power spectrum, are often of significant responses are typically considerably larger than the additional value, e.g., Fig. 3. spontaneous activity. However, the evoked response is The ability to provide real-time monitoring of each considerably smaller than the spontaneous activity in of these signal types, viz., evoked potentials (EPs), EEG, the brain. Note also that electrical noise in the op- and EMG, imposes considerable demands on the re- erating room can often swamp any neurophysiologi- sources used for transmission, display, and storage. On cal signal. the other hand, real-time EP processing requires It is evident from this discussion that to provide inter- greater computational resources but reduced use of pretable information, the neurophysiological signal must be amplified and, usually, the signal/noise other resources compared with EEG/EMG. Historically, 272 1046-2023/01 $35.00  2001 Elsevier Science All rights reserved.