A Precise Interictal Epileptiform Discharge (IED) Detection Approach Based on Transformer

Interictal epileptiform discharges(IED) are abnormal electrical discharges in the brain that play a crucial role in diagnosing epilepsy. IED detection is complex due to the non-stationary nature of electroencephalogram (EEG) signals. Meanwhile, the traditional identification of IED usually relies on manual EEG interpretation which is subjectively biased. With the development of machine learning and deep learning, computer-aided models are proposed on a fast lane in IED detection. Transformer is the latest deep learning architecture that excels at processing sequential data by employing self-attention mechanisms, enabling it to capture long-range dependencies. In this study, we proposed a novel IED detection approach, named “IED Conformer”, based on Transformer. Based on the analysis of 11 pediatric epilepsy patients, the new approach achieves an IED detection accuracy of 96.11%. The proposed method is expected to help healthcare professionals more accurately identify and manage epileptic conditions in their patients.