An intelligent model involving multi-channels spectrum patterns based features for automatic sleep stage classification. (March 2023)

• Record Type: Journal Article
• Authors: Abdulla, Shahab; Diykh, Mohammed; Siuly, Siuly; Ali, Mumtaz
• Is Part Of: International journal of medical informatics. Volume 171(2023)
• Publication Date: 2023-03
• Subjects: Sleep stages -- EEG signal -- Spectrum image -- Ensemble classifier -- MILBP; Medical informatics -- Periodicals; Information science -- Periodicals; Computers -- Periodicals; Medical technology -- Periodicals; Medical Informatics -- Periodicals; Technology, Medical -- Periodicals; Computers; Information science; Medical informatics; Medical technology; Electronic journals; Periodicals; Electronic journals; 610.285
• DOI: 10.1016/j.ijmedinf.2023.105001
• Languages: English
• ISSNs: 1386-5056

Abstract

Highlights: Robust EEG sleep classification is proposed based short-time Fourier transform. Multiple channels Local Binary Pattern is used to analyse EEG signals. An effective ensemble classification approach based genetic algorithm is designed. Abstract: Effective sleep monitoring from electroencephalogram (EEG) signals is meaningful for the diagnosis of sleep disorders, such as sleep Apnea, Insomnia, Snoring, Sleep Hypoventilation, and restless legs syndrome. Hence, developing an automatic sleep stage scoring method based on EEGs has attracted extensive research attention in recent years. The existing methods of sleep stage classification are insufficient to investigate waveform patterns, texture patterns, and temporal transformation of EEG signals, which are most associated with sleep stages scoring. To address these issues, we proposed an intelligence model based on multi-channels texture colour analysis to automatically classify sleep staging. In the proposed model, a short-time Fourier transform is applied to each EEG 30 s segment to convert it into an image form. Then the resulted spectrum image is analysed using Multiple channels Information Local Binary Pattern (MILBP). The extracted information using MILBP is then deployed to differentiate EEG sleep stages. The extracted features are tested, and the most effective ones are used to the represented EEG sleep stages. The selected characteristics are fed to an ensemble classifier integrated with a genetic algorithm which is used to select the optimal weight for each classifier, to classify EEG signal into designated sleep stages. The experimental results on two benchmark sleep datasets showed that the proposed model obtained the best performance compared with several baseline methods, including accuracy of 0.96 and 0.95, and F1-score of 0.94 and 0.93, thus demonstrating the effectiveness of our proposed model.