Evidence-based Epileptic Seizure Detection

Abstract

Epilepsy, a neurological disorder marked by recurrent seizures, arises from abnormal electrical discharges in the brain. Electroencephalogram (EEG) serves as the primary tool for detecting and diagnosing epilepsy. However, manual analysis of EEG data is both time-intensive and susceptible to human error. This study looks into the efficacy of machine learning models for automated epileptic seizure detection. We propose an Evidence-based Neural Network (ENN) for classifying epileptic seizures using EEG signals. Additionally, we introduce an uncertainty-based loss function to enhance the robustness and confidence of model predictions. The proposed approach is assessed using several performance metrics such as accuracy, precision, recall, and F1-score. Our approach demonstrated outstanding results, achieving an accuracy of 0.983 and an F1-score of 0.973. This study underscores the significance of incorporating uncertainty into model training, paving the way for reliable and precise seizure detection in clinical applications.