Deep Learning-Based Epileptic Seizure Detection from EEG Signals and PPG signals Using LSTM and CNN Models

Epilepsy is a chronic neurological disorder characterized by recurrent and unpredictable seizures that significantly affect patients' health and quality of life. Conventional diagnosis relies heavily on continuous electroencephalogram (EEG) monitoring, which requires clinical expertise and is not well suited for real-time detection. To address these challenges, this paper presents a hybrid deep learning framework that integrates Convolutional Neural Networks (CNNs) and Long Short-Term Memory (LSTM) models for automated epileptic seizure detection using multimodal EEG and Photoplethysmogram (PPG) signals. Unlike EEG-only approaches, the inclusion of PPG provides complementary physiological information such as autonomic fluctuations, seizure-induced heart rate variability changes, and peripheral vascular responses-which strengthens the model's discriminative capability, particularly in cases where EEG signatures alone are subtle or ambiguous. In the proposed framework, CNNs effectively extract spatial patterns from the preprocessed biosignals, while LSTMs capture temporal dependencies associated with seizure evolution. Data preprocessing steps including filtering, normalization, segmentation, and augmentation are applied to enhance signal quality and model generalization. The hybrid CNN-LSTM model is evaluated on benchmark EEG-PPG datasets using accuracy, precision, recall, F1-score, Cohen's Kappa, Matthews Correlation Coefficient (MCC), and Critical Success Index (CSI). Comparative analysis with existing state-of-the-art models demonstrates superior performance and robustness. Overall, the proposed multimodal deep learning system offers a reliable and efficient solution for real-time seizure detection, with strong potential for deployment in wearable and clinical healthcare platforms.