Novel Enhanced ChronoNet Model for Epileptic Seizure Detection: A Comprehensive Methodological Study

Electroencephalography (EEG) is a vital tool for capturing brain activity, playing an essential role in diagnosing various neurological disorders, particularly epilepsy. This study highlights the growing need for efficient and automated methods for seizure detection, as traditional manual analysis of EEG data can be time-consuming and prone to errors. To address this challenge, we propose two innovative approaches for classifying epileptic seizures: the enhanced ChronoNet model and EEGNet. The enhanced ChronoNet, a sophisticated recurrent neural network (RNN) architecture, integrates three convolutional layers with densely connected LSTM layers, achieving a notable 12.3% improvement in accuracy over previous versions. Meanwhile, EEGNet employs advanced deep convolutional neural network techniques for effective seizure classification. Using the CHB-MIT epilepsy dataset, we segmented the EEG recordings into 10-second epochs and rigorously compared the performance of both models. Our evaluation focused on several key metrics, including accuracy, Matthews correlation coefficient, and Cohen's kappa. The findings indicate that the enhanced ChronoNet model reaches an impressive accuracy of approximately 98.50%, underscoring its potential as a reliable solution for automated seizure detection.