Automatic Multi-label Classification of Interictal Epileptiform Discharges (IED) Detection Based on Scalp EEG and Transformer

Interictal epileptiform discharges (IED) refer to abnormal electroencephalogram (EEG) waveforms that occur between epileptic seizures, which are of great significance for the diagnosis and treatment of epilepsy. Traditionally, the detection of IED requires experienced clinical doctors to visually inspect EEG recordings, a process that is time-consuming, labor-intensive, and subject to expert bias. With the advancement of deep learning technology, computer-aided methods for automatic detection of IED have become possible, providing clinicians with faster and more accurate diagnostic tools. In this paper, we propose an automated IED detection system based on Transformer, which is capable of end-to-end identification of IED from raw EEG data. We evaluated the proposed IED detector on a dataset consisting of EEG recordings from 11 pediatric epilepsy patients collected at Wuhan Children's Hospital. The results show that the average accuracy for the multi-label classification task of different types of IED is 93.47%, and the average F1 score is 93.19%. These findings provide an effective solution for the automated detection of IED and are expected to play an important role in clinical diagnosis.