2026 International Conference on Networks Computers and Communications (ICNCCom 2026)

Epilepsy is a brain disorder with a lifetime prevalence and one of the five persistent disorders of the human nervous system. Seizures are transitory occurrences affected with extreme electrical movements in the neuron systems. Automatic seizure detection process is vital for patients through stubborn epilepsy disorder. In seizure detection, Electroencephalogram (EEG) signals are used as primary data that directly collect physiological, pathological and electrical movements of brain neurons from different patients. However, the manual interpretation of the EEG signal presents vital challenges and is essentially required for enhanced training analysis. Nowadays, Artificial Intelligence (AI) based methodology is used to increase epilepsy detection and achieve optimal performance. But, traditional Deep Learning (DL) methods have failed to capture the complex relation between EEG signals, high computational complexity and privacy limitations. So, this research develops an enhanced transformer module with a Federated Learning mechanism for enhancing Epilepsy seizure detection by EEG signals. Initially, this proposed model develops a pre-processing mechanism named the Improved Kalman-assisted Wavelet Transformer (IKassistWT) model for decreasing noisy background and capturing the effective time-series information. The extract features are passed to the Optimized Triple Attention-assisted Efficient Vision Transformer (OTEffiViT) model for learning significant features to detect epilepsy. Adjusting the hyperparameter that successfully lowers computational complexity utilizing the Oppositional Wombat Optimization Algorithm (OWOA). The weights of this local model are shared with global models, and the weights are processed using the Rényi Entropy mechanism, which incorporates extra noise to prevent patient details. The suggested model performs well in seizure detection and attains accuracy values of 99.23% as well as minimal computational complexity in experimental analysis.

Nervous Systems,Electroencephalogram,Artificial Intelligence,Federated Learning,Wavelet Transformer,Vision Transformer,Wombat Optimization