Wavelet-based Entropy Analysis on EEG Signal for Detecting Seizures

Epilepsy is one of the world's most common neurological diseases when the morbidity of brain activity occurs. People with epilepsy experience repeated seizures more than once in 24 hours. There are various ways to analyze seizures, one of them is by examining the electroencephalogram (EEG) signal recording. Neurologists usually do manual seizure analysis, yet it is time-consuming, tedious, and costive. Thus, an automated system of seizure detection is needed to assist the analysis. In this study, EEG signals of the CHB-MIT dataset are evaluated in digital signal processing using a machine learning approach. The EEG signals are filtered by applying a Butterworth bandpass filter (BPF). The mean channel selection is then applied to simplify the number of channels used in all patients. The channels are selected based on the energy value. Wavelet Packet Decomposition (WPD) is used to decompose the signals into several scales. The features are extracted using two entropy analyses, the Shannon Entropy (ShEN) and Renyi's Entropy (RE). This study used the Support Vector Machine (SVM) to classify the feature set. The channel selection process reduced the processed channels into two to five channels. The proposed method achieves the average accuracy of 85.64% and 72.48% obtained by REN and ShEN, respectively.