Self-supervised Learning with Attention Mechanism for EEG-based seizure detection

Epilepsy is a neurological disorder caused by abnormal brain discharges, which can be diagnosed by electroencephalography (EEG). Although EEG signals are usually easy to obtain, massive labeling increases the clinicians' workload. Most of the unlabeled data cannot be used directly, result in the waste of resources. We propose a self-supervised learning (SSL) method for EEG-based seizure detection. It solves the problem of insufficient annotated data by directly use large amounts of unlabeled data for training. In order to extract the global dependency of EEG signals, we apply the attention mechanism based Transformer as the backbone and name our method as Self-supervised Learning with Attention Mechanism (SLAM) for EEG-based seizure detection. Both patient-dependent and cross-patient seizure detection experiments are performed on the public CHB-MIT dataset. Experimental results verify the efficacy of SLAM.