Spatial Characteristics Analysis and Visualization for Brain Networks of Motor Imagery Based on Graph Convolutional Neural Networks

As the most complex organ in the human body, the study of the human brain has always been a hot topic in the field of neuroscience. This paper mainly studies the relevant analysis and visualization of the spatial characteristics of the brain networks on motor imagery. In view of the fact that graph convolutional neural networks can better process graph structure data in non-Euclidean space and will not lose the coupling information between EEG signal channels, they are chosen as the model structure. We also study different graph convolutional neural network structures, and find the model structure with the best classification effect by adjusting the network structure and comparing the effects of different models. In addition, we evaluated the optimal training model in many aspects. Finally, in order to observe the spatial characteristics of EEG signals more conveniently, comprehensively and from multiple levels, we designed a functional and user-friendly visualization software based on the results, which displays the diverse information of raw EEG signals and brain networks.