AUTHOR=Li Haoyang , Ji Hongfei , Yu Jian , Li Jie , Jin Lingjing , Liu Lingyu , Bai Zhongfei , Ye Chen TITLE=A sequential learning model with GNN for EEG-EMG-based stroke rehabilitation BCI JOURNAL=Frontiers in Neuroscience VOLUME=Volume 17 - 2023 YEAR=2023 URL=https://www.frontiersin.org/journals/neuroscience/articles/10.3389/fnins.2023.1125230 DOI=10.3389/fnins.2023.1125230 ISSN=1662-453X ABSTRACT=Brain-computer interfaces (BCI) have the potential to provide neurofeedback to stroke patients to help them rehabilitate motor function. However, current brain-computer interfaces often detect only rough motor intentions, ignoring the precise information about the execution of a complex movement. This is mainly because EEG signals do not contain suļ¬€icient movement execution features. This paper proposes a sequential learning model with a Graph Isomorphic Network (GIN), which takes a sequence of graph-structured data transformed from EEG and EMG. The data of each movement is divided into sub-actions and predicted separately by the model to obtain a sequential motor encoding that can reflect the sequential features of the movements. Through time-based ensemble learning, more accurate prediction results and execution quality scores can be obtained for each movement. The proposed method achieved a classification accuracy of 88.89% on an EEG-EMG synchronized dataset for push and pull movements, significantly outperforming the benchmark method of 73.23%. This approach can be used to build a hybrid EEG-EMG brain- computer interface and provide patients with more precise neural feedback to aid their recovery.