Key points are not available for this paper at this time.
According to the World Health Organization, nearly 50 million people suffer from epilepsy, one of the most common neurological disease. Epilepsy is characterized by abnormal brain activity, leading to recurrent seizures. Each seizure manifests as sudden, uncontrolled bursts of electrical activity in the brain, and the injuries and restrictions on daily life underscores the urgency of finding effective methods for epileptic seizure prediction. With the use of deep learning techniques, early predictions of epileptic seizures, an unsolved problem, are attempted in this paper. Previous research has limitations of being sensitive to noise as it is dependent on specific electroencephalogram (EEG) devices and datasets, a serious issue this paper solves. In this paper, a semi-supervised based domain generalization method to develop an accurate seizure prediction system is proposed. It consists of two phases: representation learning and transfer learning phase. To achieve high precision, the proposed method utilizes a representation learning approach. Here, a feature-swapping mechanism that effectively disentangles seizure-related features is introduced. During transfer learning, the pre-trained network is trained to output the probability of whether the input EEG indicates a seizure or not. The proposed model achieves state-of-the-art performance, with an accuracy of 90.53% and 94.88% on the NICU and Epileptic Seizure Recognition datasets respectively in within-dataset evaluations. It outperforms the previous methods by 19.35% in cross-dataset evaluations. This robust improvement opens up promising possibilities for real-world clinical applications. The proposed feature disentangling method is also expected to contribute to developing reliable medical tools.
An et al. (Thu,) studied this question.