Automated seizure detection in scalp EEG using multiple wavelet scales

The proposed research work designs a detector algorithm for automatic detection of epileptic seizures. In this work a wavelet based feature extraction technique has been adopted. Epochs of EEG are decomposed using discrete wavelet transform (DWT) up to 5 level of wavelet decomposition. Relative values of energy and a normalized coefficient of variation (NCOV) based measure, (σ²/μ_a) are computed on the wavelet coefficients acquired in the frequency range of 0-32 Hz from both seizure and non-seizure segments. The performance of NCOV over the traditionally used coefficient of variation, COV (σ²/μ²) was studied. The feature NCOV yielded better performance than the commonly used COV, σ²/μ². The algorithm was evaluated on 5 subjects from CHB-MIT scalp EEG database.