Advanced Epileptic Seizure Detection Using Deep Learning and Bayesian Optimization

Epilepsy is one of the most unpredictable and recurrent neurological disorders. For patients in rural areas, early diagnosis of epileptic seizures is crucial for timely treatment. This research uses a deep learning system to detect seizures in Electroencephalogram (EEG) signals. The dataset includes pre-seizure, seizure-free, and seizure EEG signals, and is publicly available. The study proposes using the Bayesian optimization technique with a Double LSTM model to classify EEG signals. This approach achieves up to 95% accuracy in categorizing the three types of signals, with balanced specificity and sensitivity of 96%. Feature extraction is performed using the VGG-16 model. The Double LSTM architecture, optimized with Bayesian techniques, improves classification accuracy by 6.52% compared to the SVM model and by 3.15% compared to a single LSTM model. The model's performance is evaluated using metrics such as accuracy, sensitivity, specificity, and a confusion matrix, ensuring robust classification of EEG signals into ictal, pre-ictal, and inter-ictal states. This study underscores the potential of combining deep learning with Bayesian optimization to develop effective and efficient diagnostic tools for epilepsy.