publications
publications by categories in reversed chronological order.
2025
- Blockchain-integrated multi-modal LSTM-CNN fusion for high-precision epileptic seizure detection from EEG signalsMahmoud O. Mokhiamar, Abdelrahman Mahmoud, Mohamed I. Eldagla, and 2 more authorsKnowledge-Based Systems, 2025
In the field of epileptic seizure detection, integrating advanced technologies is essential for improving accuracy and efficiency. This study introduces a pioneering approach that combines blockchain technology with a multimodal model, integrating Long Short-Term Memory (LSTM) and Convolutional Neural Network (CNN) architectures, to improve EEG-based seizure detection. The proposed approach addresses the challenge of accurate seizure detection while ensuring secure and private data management through blockchain technology. We generate spectrograms from EEG data to visualize temporal dynamics, enriching the understanding of neural activity patterns. The proposed model effectively integrates the capabilities of LSTM and CNN to capture both temporal dynamics and spatial features in the EEG data. The results show that the proposed model significantly outperforms standard LSTM and CNN models in all performance metrics, including precision, recall, accuracy, and F1-score. Notable improvements were observed, particularly in multi-class seizure detection tasks: binary (ictal vs. non-ictal), ternary (ictal, preictal, normal), and penta-class (ictal, preictal, interictal, postictal, normal). The proposed model achieved a 38.21% increase in test accuracy and a 39% increase in F1-score compared to the baseline models. For binary, ternary, and penta-class classification, the proposed model achieved accuracies of 0.991, 0.979, and 0.907, respectively. Furthermore, the integration of blockchain technology ensures secure and private data management, addressing critical concerns in medical data handling. The proposed hybrid model shows high precision, versatility, security and highlights the potential to revolutionize epileptic seizure detection and contribute to advances in the broader field of EEG analysis.