Epilepsy seizure prediction using deep learning and wavelet transform

This research focuses on predicting epileptic seizures by utilizing EEG signals through an innovative combination of Two-Dimensional Discrete Wavelet Transform (2D-DWT) for feature extraction and Convolutional Neural Networks (CNNs) for classification. The model effectively preprocesses EEG signals by removing noise and balancing data to address class imbalance. Features are extracted using 2D-DWT, which transforms EEG data into frequency-time representations, enabling the CNN to analyze spatial and temporal patterns. The CNN architecture is designed with convolutional and fully connected layers, optimized for robust classification of preictal and interictal states. The CHB-MIT scalp EEG dataset, containing recordings from pediatric patients, is used to validate the model, achieving high accuracy (98.58%), sensitivity (99.54%), and a low false positive rate (0.0153/hour). This approach demonstrates significant improvements over existing methods, offering reliable early seizure prediction to improve patient outcomes.