Cost-Effective Deep Learning Model for Seizure Detection from One Channel: Channel-Independent Classifier.

Abstract

Accurate and efficient seizure detection in epilepsy patients is a critical goal for improving their quality of life. In this study, our primary objective was to explore the feasibility of reducing the number of EEG channels required for reliable seizure detection. We accomplished this by training and verifying our model exclusively on data from the first channel of the Children's Hospital Boston dataset (CHB-MIT). Our model consistently achieved accuracies ranging from 89–99.52% across all 23 channels, demonstrating the potential for accurate seizure detection using data from a single channel. This highlights the potential for cost-effective headsets for seizure onset detection solutions in the future, which could greatly benefit individuals living with epilepsy.