Syntactic Analysis of the Electroencephalogram

Abstract
A method of quantitative electroencephalogram (EEG) analysis, based on concepts drawn from syntactic pattern recognition theories, is described. While retaining many of the quantitative characteristics of traditional EEG analysis methodologies, the system also utilizes the spatiotemporal characteristics of the EEG record in a fashion similar in concept to visual EEG scoring. Single channels of multichannel EEG's are sequentially partitioned into 100 1-s epochs, and each epoch is assigned, by discriminant analysis, one of seven possible labels. The concatenation of these labels defines an "EEG sentence" which is parsed using a context-free grammar into increasingly abstract, yet clinically meaningful, "EEG phrases." Interchannel differences and similarities are evaluated after the parse using a cross-channel contextual analyzer. System performance is described, and examples of parse results are given for normal and abnormal EEG's.

This publication has 9 references indexed in Scilit: