Intraoperative EEG patterns: implications for EEG monitoring

Univariate descriptors such as the mean frequency and spectral edge frequency have been proposed for intraoperative representation of the EEG. Such univariate descriptors of the EEG are accurate only when the EEG behaves as a unimodal distribution of frequencies that change slowly with time. EEGs were recorded from 64 patients undergoing anesthetic inductions and 30 patients undergoing cardiopulmonary bypass to determine the characteristics of the observed distribution of frequencies. Multimodal EEG activity was observed in 64% of these cases, including 83% of those patients undergoing cardiopulmonary bypass. The differences between the two peaks averaged 7.6 Hz, and the average ratio of the power of the peaks to the intervening valley was 2.5:1 and 1.9:1. Calculations of mean frequency and spectral edge frequency failed to adequately reflect the complexity of the EEG in these cases. Burst-suppression activity was observed in 26% of cases during cardiopulmonary bypass, and averaging over time destroyed the characteristic pattern. Thus, univariate descriptors of the EEG appear inadequate to describe the behavior of EEG during anesthesia in a large percentage of cases.