Detection of Epileptic Seizures Based-on Channel Fusion and Transformer Network in EEG Recordings

• DOI: 10.14569/ijacsa.2023.01405110
• Publication Date: 2023-06-05T08:25:28Z
• Authors: Jose Yauri; Manuel Lagos; Hugo Vega-Huerta; Percy De-La-Cruz-VdV; Gisella Luisa Ele...; Enrique Condor-Ti...

According to the World Health Organization, epilepsy affects more than 50 million people in world, and specifically, 80% of them live developing countries. Therefore, has become among major public issue for many governments deserves be engaged. Epilepsy is characterized by uncontrollable seizures subject due a sudden abnormal functionality brain. Recurrence attacks change people's lives interferes with their daily activities. Although no cure, it could mitigated an appropriated diagnosis medication. Usually, based on analysis electroencephalogram (EEG) patient. However, process searching seizure patterns multichannel EEG recording visual demanding time consuming task, even experienced neurologists. Despite recent progress automatic recognition epilepsy, nature recordings still challenges current methods. In this work, new method detect proposed. First, uses convolutions perform channel fusion, next, self-attention network extracts temporal features classify between interictal ictal states. The was validated CHB-MIT dataset using k-fold cross-validation achieved 99.74% specificity 99.15% sensitivity, surpassing approaches.