JoWUA

Abstract

Epileptic seizure detection using EEG signals is one of the most vital problems in biomedical and ubiquitous health care systems, since it is a very important process for real-time monitoring and decision making. But EEG signals are very much non-stationary and noisy, and thus, their detection becomes a difficult process. In this paper, a bio-inspired optimized hybrid Convolutional Neural Network-Recurrent Neural Network (CNN-RNN) approach is used to make seizure detection efficient and accurate. The CNN takes care of the spatial features in the EEG signal, whereas the RNN takes care of temporal dependencies in sequence learning. Moreover, a bio-inspired optimized technique is applied to the hybrid model in order to optimize learning rate, filter size, kernel size, and hidden units. The proposed algorithm will be validated using EEG benchmark data sets such as CHB-MIT and the Bonn University dataset. The experimental outcomes show that the suggested technique provides high classification accuracy of 96.8%, precision of 95.9%, recall of 96.2%, and F1 score of 96.0%. This indicates the superiority of the proposed technique over traditional techniques, including SVM, CNN, LSTM, and CNN-LSTM. Also, the false alarm rate becomes 2.9%, and detection latency is reduced, which makes the proposed model fit for real-time applications. The experimental results highlight the effectiveness of bio-inspired optimization and hybrid deep learning. The suggested technique is appropriate for deployment in wireless mobile/wearable healthcare systems for real-time and effective seizure detection and monitoring. Further research will be devoted to the development of a lightweight model, federated learning, and edge computing.