Event-Based Epileptic Seizure Detection with Stacked Convolutional Restricted Boltzmann Machine

Abstract

Research indicates that around 1% of the global population experiences epileptic seizures, described as an undue neuronal discharge in the brain, which can significantly impact the quality of life for those affected. Children who experience frequent or prolonged seizures may experience lasting cognitive damage and psychological difficulties, and sudden seizures can pose a serious risk of injury or even death. Therefore, developing an accurate machine learning model to predict seizures is crucial. This study introduces an Event-based Stacked Convolutional Restricted Boltzmann Machine architecture that can detect ictal periods in EEG signals. The model was trained and evaluated on a publicly available dataset of EEG recordings from 24 pediatric patients, achieving a high prediction accuracy of 94.2%, with a mean prediction time of 19.62 min. The model's mean sensitivity and specificity rates were also comparable to those of state-of-the-art predictive models on the CHB-MIT dataset.