Automated EEG seizure detection based on S-transform

Epilepsy detection using EEG signals is an important clinical practice to study the occurrence of seizures. There is a need to analyze huge volumes of EEG data for finding the epileptic seizures. The manual analysis of EEG records for identifying seizure manifestations is time-consuming and creates an immense workload for the physician. To reduce the EEG analysis time, an autonomous epilepsy detection system is proposed using a Time-Frequency(TF) entropy measure. First, the TF spectrum of the EEG signal is computed using the S-transform(ST). Then the entropy measure is determined from the TF spectrum. The performance metrics of the proposed entropy index is measured using a least square support vector machine(LSSVM) classifier. The proposed entropy feature produced a highest classification accuracy of 86% when validated with Bern-Barcelona EEG dataset. The area under curve(AUC) of the receiver operating characteristic(ROC) plot for the proposed entropy feature is 0.914. The average computational time for extracting the proposed entropy feature is 0.4027s. The proposed Time-Frequency entropy feature is analyzed in terms of classification accuracy and computation time.