DST-GNN: A dynamic spatio-temporal graph neural network for motor imagery classification

Electroencephalography (EEG)-based motor imagery classification plays an important role in brain-computer interface (BCI) systems. However, existing methods often struggle to effectively capture the complex spatial and temporal dependencies among EEG channels and usually rely on manually designed prior knowledge. To address these limitations, this paper proposes a Dynamic Spatial-Temporal Graph Neural Network (DST-GNN) for motor imagery classification. The proposed framework models EEG signals as dynamic graphs and jointly learns spatial interactions and temporal patterns from multi-channel EEG data. In addition, a graph readout mechanism is employed to generate hierarchical spatial-temporal representations, enabling more comprehensive feature aggregation for classification. Extensive experiments conducted on a public motor imagery dataset demonstrate that DST-GNN consistently outperforms representative baseline methods and achieves competitive classification performance.