EEG-based epileptic seizures detection with adaptive learning capability

Abstract

Epilepsy is considered one of the most common neurological disorders. Epileptic seizures can be a major life disability that might result in loss of consciousness, and/or injury to oneself or others. This research work aims to develop an epileptic seizure detection method using electroencephalography (EEG) signal analysis. We combine discrete wavelet transform (DWT), Shannon entropy, and statistical feature (standard deviation) to extract distinctive features of a given EEG signal. K-nearest neighbors (KNN) automatically classifies the EEG signal by comparing the extracted features with the features of the normal and seizure baseline. Adaptive learning is used to continuously update the two baselines based the user feedback, if needed, to improve its performance over the time. Our proposed method achieved an overall 94.5 % sensitivity tested with up to 570 hours of continuous EEG recording from ten patients with total of 55 seizure events taken from CHB-MIT database. With its simplicity and fast processing time, the proposed method is suitable to be implemented in embedded device which has limited processing resource.