Brain state identification and forecasting of acute pathology using unsupervised fuzzy clustering of EEG temporal patterns

The electroencephalogram (EEG) signal, being the superficially recorded gross electrical activity of the brain, is a non-stationary, continuously fluctuating signal, characterized both by the frequency distribution of its ongoing background pattern and by the existence and form of single waves or complexes of physiological or pathological origin. Both characteristics are, on the one hand, state specific and, as such, amenable to classification by brain state but, on the other hand, possess enough variability, overlap, and vague transition to require fuzzy classification. In addition, the number of underlying semi-stationary states or processes in the continuously sampled signal is both unknown and time-varying, a fact that requires an adaptive selection of the number of classes.