Wavelet based approach for detecting and classifying epileptiform waves in EEG

The automatic detection and classification of epileptiform waves in EEG has great clinical significance. A hierarchical approach for achieving this task is presented in this paper which is based on time-scale analysis of the EEG. With 16-channel input EEG data, a preliminary screening is first done to pick out segments suspicious of containing epileptic waves. It is based on an adaptive filtering process to separate the non-stationary transients from the stationary background EEG. These segments are then decomposed into different scales by applying the continuous wavelet transform. After feature extraction, recognition and classification are conducted based on the combination of a fuzzy neural network and rule-based expert system. Finally, a statistical analysis of the obtained results is made and reported. The initial clinical results are encouraging.