Automatic epilepsy detection based on generalized convolutional prototype learning. (November 2021)

• Record Type: Journal Article
• Authors: Lyu, Chengang; Chen, Yuxin; Chen, Zhijuan; Liu, Yuheng; Wang, Zengguang
• Is Part Of: Measurement. Volume 184(2021)
• Publication Date: 2021-11
• Subjects: Classification -- EEG -- Epilepsy detection -- Generalized convolutional prototype learning (GCPL) -- Clinical practice -- Robust; Weights and measures -- Periodicals; Measurement -- Periodicals; Measurement; Weights and measures; Periodicals; 530.8
• DOI: 10.1016/j.measurement.2021.109954
• Languages: English
• ISSNs: 0263-2241

Abstract

Highlights: An epilepsy detection method based on generalized convolutional prototype learning (GCPL) is used to classify EEG signals. GCPL combines convolutional neural network (CNN) and prototype learning. GCPL maps EEG samples to the regions close to the prototype in the feature space. The introduction of prototype loss into CNN and GCPL can reduce the over-fitting phenomenon. Abstract: In clinical applications, different EEG acquisition equipment and experience of neurologists make the quality and pattern of EEG signals different. This hinders the promotion of some automatic epilepsy detection methods in clinical practice. In order to solve this problem, we proposed an epilepsy detection method based on generalized convolutional prototype learning (GCPL). GCPL maps EEG samples to the regions close to the prototype in the feature space, which is more suitable for the classification of clinical EEG signals with diverse quality and pattern. Moreover, it can overcome the impact of the change of discriminative features caused by different datasets on the classification performance, which shows more robustness. In the experiments, the proposed method could reach a sensitivity of 98.75%, a specificity 100% of and an accuracy 99.38%, respectively on the clinical dataset. GCPL can complete a test within 2 s, which meets the actual clinical needs.