Modified ResNet50 with Transfer Learning for Enhanced Epileptic Activity Detection Using EEG Data

Epilepsy is a neurological disorder that significantly impacts individuals globally, particularly through recurrent seizures. Electroencephalograms (EEGs) are a crucial tool for diagnosing epilepsy, but manual analysis of EEG data is both labor-intensive and prone to errors. To address this, we propose an automated epilepsy detection system based on deep learning. Our approach utilizes a pre-trained ResNet50 model, specifically modified to accommodate the unique characteristics of EEG data through transfer learning. Evaluated on the CHB-MIT dataset, our model achieved an accuracy of 97% on test data, demonstrating its ability to effectively generalize epilepsy-related patterns. This provides a reliable and efficient method for EEG-based seizure detection.