Construction and Analysis of Cortico-Muscular Coupling Networks Using Transfer Entropy Computed After High-Dimensional Time-Delay Embedding

Cortico-muscular coupling reflects the interaction between the cerebral cortex and muscles during motor control. This study proposes a transfer entropy model based on time-delay mutual information and high-dimensional signal embedding (HDEmbedDMI-TE) to construct cortico-muscular coupling networks. Electroencephalography and electromyography were recorded from healthy participants and stroke patients during a five-finger grasping (FG) task. After preprocessing, HDEmbedDMI-TE was used to quantify the direction and strength of information flow between cortical regions and muscles. Simulation experiments with known causal directions validated the proposed method, demonstrating accurate detection of directional coupling and a $\mathbf{4. 8}$-fold improvement in computational efficiency compared with conventional Transfer Entropy. Network analysis revealed significant directional coupling during FG, with information predominantly flowing from the motor cortex to muscles in healthy individuals, whereas stroke patients showed impaired motor cortical regulation and compensatory involvement of cognitive-related regions. These results demonstrate the effectiveness of HDEmbedDMI-TE for corticomuscular coupling analysis and its potential value for motor control research and neurorehabilitation assessment. This work advances signal processing methodologies for multimodal data through efficient nonlinear coupling and network analysis.