Automated detection of epileptiform transients in the EEG using a multi-channel expert system approach

A PC-based system that has been developed for detection of interictal epileptiform events in the EEG is discussed. A feature extraction stage inspects each EEG channel for spikelike occurrences on the basis of parameters such as duration, amplitude, and sharpness. An expert system stage then examines each of the possible spikes from a multichannel perspective to decide whether they constitute a true epileptiform event. The results of a preliminary pilot trial of the PC-based system are reported. In an analysis of EEGs from five patients, epileptiform events were detected in all EEGs with epileptiform activity. Most missed events were due to failure to detect sufficient coincident spikes in the first pass because the spikes were of very low amplitude or there were bursts of muscle artifact. There were no false detections.<>