Analysis of EEG changes with carotid clamping by selective analog filtering, matched inverse digital filtering and automatic adaptive segmentation: A comparative study

Abstract

A comparative evaluation was made of 3 methods for possible use in a system for automatic evaluation and signaling of significant EEG changes associated with carotid clamping. The methods were: (1) selective analog filtering in two frequency bands (‘fast’ — 4–20 c/sec; and ‘slow’ — 0.8–1.0 c/sec); (2) inverse digital filtering, a computer-based method in which a filter is formed from a pre-clamping EEG baseline such that only frequency components different from the baseline components appear in the filter output; and (3) automatic adaptive segmentation, a computer-based method of detecting and signaling EEG changes by comparison of the autocorrelogram of a pre-clamping reference sample with that from an equal-duration sample ‘seen’ through a progressively moving window in the post-clamping period. For the analyses, portions of ink-recorded monitoring EEGs previously interpreted as showing changes consequent to carotid clamping were transferred to a cassette tape recorder system by means of a multichannel photo-optical scanning system. Of the 3 methods, automatic adaptive segmentation was clearly the best. For selective analog filtering, the thresholds for a significant change were difficult to set. Inverse digital filtering was unsatisfactory because of its insensitivity to changes in the amplitude frequency components already present, although new frequency components could be detected. An incidental finding was that a slow-speed write-out of the analog filter outputs provided a good visual indication of EEG changes.