Enhancing attention network spatiotemporal dynamics for motor rehabilitation in Parkinson’s disease

Recent research has achieved significant advances in multimodal neuroimaging technologies for objective assessment of motor rehabilitation efficacy in Parkinson's disease (PD). Nevertheless, rapid and precise identification of treatment-responsive patients remains challenging, often requiring lengthy clinical evaluations and lacking neural biomarkers. By integrating electroencephalography (EEG) microstate analysis with functional magnetic resonance imaging (fMRI) coactivation patterns (CAPs), we enable spatiotemporal characterization of attention network reorganization for predictive stratification. "The multimodal approach leverages microstate dynamics and CAP dwell time to identify responders with 86% accuracy before intervention," explained study author Tianyi Yan. The platform combines (a) EEG microstate sequencing to quantify sensory-cognitive network transitions, (b) fMRI CAP analysis to map attentional resource allocation, and (c) machine learning integration for clinical outcome prediction. "This solution provides a neurophysiological framework to optimize rehabilitation resource allocation, reducing assessment time from months to minutes," emphasized the authors. Thus, they developed a biomarker-driven protocol comprising resting-state EEG/fMRI acquisitions, validated against motor function scales and quality-of-life metrics.

EEG microstate analysis exploits millisecond-scale topographical fluctuations to map large-scale neural dynamics. Four canonical classes (A-D) represent distinct network states: microstate D (frontal-centric) correlates with dorsal attention network (DAN) engagement, while microstate C (anterior-posterior) reflects default mode network (DMN) activity. "The transition probability from sensory microstates (A/B) to attentional microstate D significantly increases in responders (P<0.01)," validated by CARTOOL clustering. Post-MIRT, responders exhibit decreased microstate C occurrence (−28.7%) and increased microstate D (+31.2%), contrasting nonresponders' inverse patterns—confirmed via global explained variance analysis. This network shift underpins motor recovery.

Concurrently, fMRI CAPs capture transient network coactivation through sliding-window correlation. Fourteen CAPs were identified, with CAP (DAN⁺/VAN⁻) showing prolonged dwell time in responders (P=0.012). "The antagonistic DAN-VAN dynamics inversely correlate with microstate C suppression (ρ=−0.315, P=0.048)," demonstrated by GRETNA-based k-means clustering. When integrated with EEG microstate parameters, the SVM classifier achieves 86% accuracy in predicting MIRT response, outperforming clinical scales alone (50.7%). Specificity was confirmed using PDQ-39 quality-of-life scores at 3-month follow-up. "The modular design allows adaptation to diverse rehabilitation protocols by retraining the classifier," enabling personalized treatment planning. Compared to conventional 4-week rehabilitation programs, this approach reduces intervention time to 2 weeks while maintaining efficacy. However, current limitations include medication confounding effects and lack of healthy control baselines. Future work will focus on longitudinal validation and integration with closed-loop neuromodulation. Collectively, this neuroimaging framework offers a transformative tool for precision rehabilitation in PD, advancing resource optimization in neurodegenerative care.

Authors of the paper include Guangying Pei, Mengxuan Hu, Jian Ouyang, Zhaohui Jin, Kexin Wang, Detao Meng, Yixuan Wang, Keke Chen, Li Wang, Li-Zhi Cao, Shintaro Funahashi, Tianyi Yan, and Boyan Fang.

This work was supported by the National Natural Science Foundation of China (grant numbers 82202291 and 62336002), the Beijing Natural Science Foundation (grant numbers 7242274 and S23114), the Key-Area Research and Development Program of Guangdong Province (grant number 2023B0303030002), the STI 2030-Major Projects (grant number 2022ZD0208500), the Beijing Nova Program (grant number 20230484465), and the Science and Technology Development Fund of Beijing Rehabilitation Hospital, Capital Medical University (grant number 2023R-04).

The paper, “Enhancing Attention Network Spatiotemporal Dynamics for Motor Rehabilitation in Parkinson’s Disease” was published in the journal Cyborg and Bionic Systems on Jun 19, 2025, at DOI: 10.34133/cbsystems.0293.