An Epileptic EEG Classification Approach with Spike Train Encoding Using Spiking Neural Networks

Accurate recognition of epileptic electroencephalogram (EEG) signals holds significant practical implications for diagnosis and treatment of epilepsy. Conventional machine learning classification methodologies heavily rely on feature extraction and selection techniques for epilepsy EEG data, potentially leading to the loss of valuable information pertinent to epilepsy classification. Utilizing the spike train encoding scheme to convert the EEG data, bypassing the need for feature extraction, this paper presents an approach for classifying epileptic EEG using spiking neural network model, which is trained by a novel supervised learning algorithm based on the desired spike train reconstruction process. The original EEG data is used to generate spike trains through BSA encoding method. Then, the epileptic classification model based on multilayer feedforward spiking neural networks is developed for the detection of epileptic seizures. The proposed classification approach achieved good performance on CHB-MIT scalp EEG benchmark dataset.