Detecting transient cognitive impairment with EEG pattern recognition methods.

Abstract

This paper describes an initial evaluation of a new method for assessing transient states of cognitive impairment associated with intoxication or fatigue: neural network pattern recognition applied to features of the electroencephalogram (EEG) recorded from subjects performing a standardized task. Nine subjects performed a working memory task during an extended testing session occurring over the course of one night, and encompassing an alert baseline period, a state of mild acute intoxication, and a state of fatigue compounded by "hangover" or intoxication after-effects. Relative to the alert baseline, task performance was less accurate in the other test conditions, providing evidence of transient cognitive impairment. These states of impairment were associated with changes in spectral characteristics of the EEG. Neural network-based EEG pattern recognition techniques were used to develop and test detectors of these changes. Brief testing data samples originating from the alert baseline condition could be discriminated from those recorded during the state of acute intoxication with 98% accuracy (p < 0.0001), and from those recorded during the state of fatigue/hangover with 92% accuracy (p < 0.001). Furthermore, networks trained on data from a group of subjects were found to accurately classify data from test subjects who were not part of the training group. These results demonstrate the feasibility of using neurophysiological monitoring methods for detecting transient cognitive impairment.