EEG During Anesthesia Is Not a Linear Random Process

Abstract

To the Editor: We read the recent article written by Jeleazcov et al. (1) with great interest. They analyzed 90 h of electroencephalogram (EEG) recorded from 20 patients. They found that the bicoherence was zero or a mere constant. We found a significant increase of EEG bicoherence during isoflurane or sevoflurane anesthesia (2–4). We also found the same changes in EEG bicoherence during propofol-fentanyl anesthesia (data not shown). What is the difference? Bispectral analysis quantifies the phase relation (quadratic phase coupling) among the frequency components of a signal in stationary state. This means that all epochs used for calculation of bispectrum/bicoherence must be taken from a signal in the same stationary state, whereas Jeleazcov et al. (1) calculated EEG bicoherence from EEG data recorded from different patients in various anesthetic states, which is theoretically violent. We could not get correct EEG bicoherence values with such a method. We should analyze the EEG recorded at one anesthetic state of one patient. Although we could occasionally obtain meaningful EEG bicoherence data from EEG in transient state (2), they were probably exceptional cases. We think that EEG during anesthesia cannot be a linear random process. Satoshi Hagihira, MD, PhD Department of Anesthesiology Osaka University Graduate School of Medicine [email protected] Masaki Takashina, MD Surgical Center, Osaka University Hospital Takahiko Mori, MD Department of Anesthesiology Osaka General Medical Center Takashi Mashimo, MD, PhD Department of Anesthesiology Osaka University Graduate School of Medicine Osaka, Japan