学术文献库

Burst-suppression (BS) is a reliable electroencephalogram (EEG) indicator of excessive deep anesthesia common in mammals. Since some intermittent events are known to follow a power-law, we investigated the power-law hypothesis in BS by comparing it with alternative functions focusing on flattish periods and developed a new method for detecting BS as an application of statistical model in dogs. Young-adult 6 beagles and senior 6 beagles were anesthetized with sevoflurane 2.0%-5.0% and three of 64 sec EEG (256 Hz) from Fpz-T4 via scalp electrodes were recorded. Three thresholds for peak-to-peak voltage were set: mean value of peak-to-peak voltage at sevoflurane 2.0% in each dog (AS2%), 3mcrV, and 5mcrV. The subthreshold periods were discriminated as $τ$ events. We fitted the empirical probability distribution of $τ$ by a power-law distribution and an exponential distribution. These two distributions were compared by the normalized log-likelihood ratio test to see which distribution was better fit. At sevoflurane 2.0%-3.0%, by any threshold, the exponential distribution became better fit in all dogs. The power-law distribution became better fit only when BS expressed on EEG. No strict threshold was required for detection of onset of BS. We showed a transition from exponential behavior to power-law behavior on the right tail of $τ$ distributions in response to the appearance of suppression waves with increasing anesthetic. This will be a robust tool for BS detection.