EEG-fNIRS multilayer brain network analysis revealed functional neural reorganization of rTMS with motor training in stroke

• Published in: IEEE transactions on biomedical engineering Vol. PP; pp. 1 - 12
• Main Authors: Lin, Jianeng, Jin, Shuxiao, You, Yugen, Liu, Jinrui, Lu, Jiewei, Shu, Zhilin, Feng, Yuxin, Zhang, Yaru, Xiao, Hui, Zhang, Ying, Wang, Jing, Zhao, Xintong, Wang, Chunfang, Han, Jianda, Yu, Ningbo
• Format: Journal Article
• Language: English
• Published: United States IEEE 18.06.2025
• Subjects: Artificial intelligence; Couplings; EEG-fNIRS; Electroencephalography; Functional magnetic resonance imaging; Functional near-infrared spectroscopy; Hands; Motors; multilayer brain network; neurovascular coupling; Nonhomogeneous media; rTMS; Stroke; Stroke (medical condition); Training
• ISSN: 0018-9294; 1558-2531; 1558-2531
• DOI: 10.1109/TBME.2025.3580943

Objective: Repetitive transcranial magnetic stimulation (rTMS) is an effective non-invasive neuromodulation technique promoting motor function recovery in stroke patients. Our study aimed to reveal the functional neural reorganization of rTMS with motor training in stroke from a comprehensive multimodal perspective. Methods: This study proposed a novel EEG-fNIRS multilayer brain network analysis method to investigate the hemisphere activation and neuroplasticity changes and conducted clinical study. Specifically, the EEG-fNIRS signals were first reconstructed and aligned in the unified cortical source space. Then, the neurovascular coupling strength was quantified by subject-specific estimation of the hemodynamic response function and utilized to build the interlayer edges. Subsequently, the unimodal intra-layer edge and bimodal inter-layer edge were combined to construct the multilayer brain network, of which features were extracted. 27 stroke patients and 13 healthy controls were recruited in the clinical experiment. Results: We found that the rTMS group showed significant improvement in the neurovascular coupling levels and multiplex clustering coefficients compared with the sham group. Moreover, these neural changes were significantly correlated with the motor function improvements (R = 0.600 and 0.618). The proposed method reduces the prediction error for rehabilitation outcomes by an average of 20.36% compared to unimodal approaches. Conclusion: The results indicated that our method effectively reveals the functional neural reorganization of rTMS with motor training in stroke. Significance: This work provides a novel method to empower neuroelectric-hemodynamic analysis and a unique insight into the mechanisms of stroke recovery and the therapeutic potential of rTMS in combination with motor training.