Computerized analysis of epileptic activity and sleep in mobile long-term EEG monitoring

Implemented on a LSI 11/73 microcomputer, an approach is presented to perform pattern recognition of one channel in high speed (factor 20). A mimetic algorithm, acting in time domain, continuously tests the signal with respect to the occurrence of spike-wave-like patterns and calculates a low frequency index describing the time of sleep. Those epochs suspected of containing paroxysmal activity are stored on disk for subsequent editing. In a similar way, 8-channel analysis is possible if the replay speed is less than 20 times real time.