An Approach to Seizure Onset Detection Using Fuzzy Logic Based on Seizure Evolution in Intracranial EEG

We present a fuzzy rule-based system for epileptic seizure onset detection. Two features based on temporal evolution of seizure in electroencephalogram (EEG) were extracted from intracranial EEG (iEEG) recordings. Features extracted from multichannel EEGs were combined using fuzzy algorithms in feature domain as well as in spatial (channels) domain. Fuzzy rules were derived from experts’ knowledge and reasoning. Finally, a predefined threshold was used to make the final decision. A total of 40.46 h of iEEG recordings (obtained from Freiburg Seizure Prediction EEG database) selected from 13 patients having 19 seizures was used for the system evaluation. The overall detection rate of 100% was achieved with false detection rate of 0.275/h and the average detection latency of 26.858 seconds.