Epileptic Seizure Detection Using CWT Scalograms and Deep Convolutional Neural Networks

Epilepsy is a neurological disorder characterized by recurrent seizures, affecting millions globally. Electroencephalography (EEG) is a critical diagnostic tool, but manual interpretation is time-consuming and subjective. This study proposes an automated seizure detection system that transforms EEG signals into time-frequency scalograms using Continuous Wavelet Transform (CWT), followed by classification through transfer learning with deep Convolutional Neural Networks (CNNs). Preprocessing includes artifact rejection, noise filtering, and statistical channel selection. EEG recordings are segmented into 5-second windows to preserve both spectral and temporal features. Several CNN architectures—including VGG16, ResNet50, InceptionV3, MobileNetV2, and a customized SE–CNN—were evaluated using the TUH EEG Corpus. The proposed SE–CNN model achieved a test accuracy of 99.29%, with precision of 100%, recall of 98.59%, and F1-score of 99.29%, outperforming several recent deep learning approaches that achieved accuracies in the range of 87-98.7%. These results demonstrate the system's strong potential for real-time clinical application and remote epilepsy monitoring.