Enhanced Epileptic Seizure Detection from Electroencephalogram Signals Using Convolutional Neural Networks

This is the authors' abstract. We don't add key points for this paper.

Epilepsy affects millions of people globally and is characterized by unpredictable, recurring seizures. Although electroencephalography (EEG) is essential for diagnosing and tracking these events, relying on human experts to review the recordings is Manual examination of EEG recordings requires substantial time and may lead to variations in interpretation. To overcome these obstacles, this research introduces a fully automated seizure detection framework powered by artificial neural networks. Using the widely recognized Bonn University EEG dataset, we differentiate between seizure and nonseizure signals. Rather than relying on manual feature engineering, our proposed convolutional neural network identifies distinctive patterns directly from the raw, preprocessed EEG time-series.The proposed architecture is assessed through commonly used classification measures, including accuracy, precision, sensitivity, and F1-score. Our results confirm that this deep learning approach can accurately and robustly separate epileptic events from normal brain activity. By reducing the need for manual expert review, this computer-aided diagnostic tool shows strong potential for real-time clinical use and continuous, efficient patient monitoring.