Graph Convolution Neural Network Based End-to-End Channel Selection and Classification for Motor Imagery Brain–Computer Interfaces

Abstract

Classification of electroencephalogram-based motor imagery (MI-EEG) tasks is crucial in brain-computer interface (BCI). EEG signals require a large number of channels in the acquisition process, which hinders its application in practice. How to select the optimal channel subset without a serious impact on the classification performance is an urgent problem to be solved in the field of BCIs. This article proposes an end-to-end deep learning framework, called EEG channel active inference neural network (EEG-ARNN), which is based on graph convolutional neural networks (GCN) to fully exploit the correlation of signals in the temporal and spatial domains. Two channel selection methods, i.e., edge-selection (ES) and aggregationselection (AS), are proposed to select a specified number of optimal channels automatically. Two publicly available BCI Competition IV 2a (BCICIV 2a) dataset and PhysioNet dataset and a self-collected dataset (TJU dataset) are used to evaluate the performance of the proposed method. Experimental results reveal that the proposed method outperforms state-of-the-art methods in terms of both classification accuracy and robustness. Using only a small number of channels, we obtain a classification performance similar to that of using all channels. Finally, the association between selected channels and activated brain areas is analyzed,

Investigating the Impact of Guided Imagery on Stress, Brain Functions, and Attention: A Randomized Trial

“…Moreover, a notable feature of this research involved the application of multi-sensor EEG signal classification and a GLM for the categorization of two mental states. These findings offer compelling evidence regarding the potential for developing innovative therapies in the domain of human-machine interactions like in [103] and that EEG is not the only medium that can be used to support human-machine interaction control [104,105]. For instance, the study titled "Golden Subject Is Everyone: A Subject Transfer Neural Network for Motor Imagery-based Brain Computer Interfaces" [106] explores the use of neural networks to transfer knowledge between individuals in the context of motor-imagery-based brain-computer interfaces.…”

Gender difference in functional activity of 4-months-old infants during sleep: A functional near-infrared spectroscopy study

“…A study proposes an end-to-end deep learning framework called EEG channel active inference neural network to handle the classification of electroencephalogrambased motor imagery (42). Regarding the results obtained in subsection 3.3 "Significance tests for individual channels, " in conjunction with this one study, we suggest that the differences in HbO may be due to differences in the timing or degree of development of the frontoparietal network (CH1), somatomotor network (CH32, 39) and the visual network (CH21) in male and female infants (42,43). For the HbR differences, we suggest that the differences may be due to differences in the timing or extent of development of the frontoparietal network (CH1), somatomotor network (CH9, 15), and the dorsal network (CH6) (42,43).…”

“…In the last decade or so, more and more cognitive neuroscientists have shifted their focus from isolated “regions” to larger “networks.” A study proposes an end-to-end deep learning framework called EEG channel active inference neural network to handle the classification of electroencephalogram-based motor imagery ( 42 ). Regarding the results obtained in subsection 3.3 “Significance tests for individual channels,” in conjunction with this one study, we suggest that the differences in HbO may be due to differences in the timing or degree of development of the frontoparietal network (CH1), somatomotor network (CH32, 39) and the visual network (CH21) in male and female infants ( 42 , 43 ). For the HbR differences, we suggest that the differences may be due to differences in the timing or extent of development of the frontoparietal network (CH1), somatomotor network (CH9, 15), and the dorsal network (CH6) ( 42 , 43 ).…”

Dynamic pruning group equivariant network for motor imagery EEG recognition

“…With the deepening of the application of the deep learning method in the field of EEG signal processing, the end-to-end learning method combining EEG feature extraction and classification shows obvious advantages. In recent years, many EEG classification methods based on the deep learning model show superior performance to traditional methods ( Ek and Bma, 2021 ; Idowu et al, 2021 ; Sun et al, 2020 ; Sun et al, 2022a ; Yang et al, 2015 ). As the first deep learning model introduced into EEG signal processing, the CNN integrates EEG feature extraction and classification and has achieved good final classification results.…”

“…As the first deep learning model introduced into EEG signal processing, the CNN integrates EEG feature extraction and classification and has achieved good final classification results. At the same time, graph convolutional network and transfer learning technology have also been introduced into brain computer interface research, and some new progress has also been made ( Zhang et al, 2021 ; Sun et al, 2022b ; Sun et al, 2023 ).…”