Application of brain-computer-interface in awareness detection using machine learning methods

• Published in: China communications Vol. 19; no. 6; pp. 279 - 291
• Main Authors: Feng, Kaiqiang, Lin, Weilong, Wu, Feng, Pang, Chengxin, Song, Liang, Wu, Yijia, Cao, Rong, Shang, Huiliang, Zeng, Xinhua
• Format: Journal Article
• Language: English
• Published: China Institute of Communications 01.06.2022; School of Information Science and Technology,Fudan University,Shanghai 200433,China%Academy for Engineering and Technology,Fudan University,Shanghai 200433,China%Department of Cardiology,Yueyang Hospital of Integrated Traditional Chinese and Western Medicine,Shanghai 200437,China%Electronics and Information Engineering,Shanghai University of Electric Power,Shanghai 200090,China%School of Information Science and Technology,Fudan University,Shanghai 200433,China; Academy for Engineering and Technology,Fudan University,Shanghai 200433,China
• Subjects: Auditory system; awareness detection; brain-computer interface; deep learning; EEG; Electroencephalography; Functional magnetic resonance imaging; machine learning; Presses; Task analysis; Temporal lobe; Visualization; deep learning; machine learning; brain-computer interface; EEG; aware-ness detection
• ISSN: 1673-5447
• DOI: 10.23919/JCC.2022.06.020

The awareness detection in patients with disorders of consciousness currently relies on behavioral observations and CRS-R tests, however, the mis-diagnosis rates have been relatively high. In this study, we applied brain-computer interface (BCI) to awareness detection with a passive auditory stimulation paradigm. 12 subjects with normal hearing were invited to collect electroencephalogram (EEG) based on a BCI communication system, in which EEG signals are transmitted wirelessly. After necessary preprocessing, RBF-SVM and EEGNet were used for algorithm realization and analysis. For a single subject, RBF-SVM can distinguish his (her) name stimuli awareness with classification accuracies ranging from 60-95%. EEGNet was used to learn all subjects' data and improved accuracy to 78.04% for characteristics finding and model generalization. Moreover, we completed the supplementary analysis work from the time domain and time-frequency domain. This study applied BCI communication to human awareness detection, proposed a passive auditory paradigm, and proved the effectiveness, which could be an inspiration for brain, mental or physical diseases diagnosis and detection.