A knowledge-based approach to the automated detection of EEG waveforms

Reliable automated detection of EEG waveforms should involve adequate and efficient utilization of heuristics used in the visual EEG analysis process, which include spatiotemporal information. The knowledge-based approach to automated EEG waveform detection quantifies efficiently the highly qualitative reasoning used in visual EEG analysis and provides an effective means for dealing with the problem of misses and false positive detections. The knowledge-based system follows an object-oriented strategy with procedural information attached to frame-based descriptions. It is also implemented in KEE on a Symbolics 3640 LISP computer, and involves a blackboard scheme as well. The overall system has been tested with data from several subjects, and it shows good agreement with the electroencephalographer minimizing false positive detections due to sharp but normal EEG activity, or to sharp but artifactual (non-EEG) activity.<>