Source-Level Resting-State EEG Connectivity Reveals Frequency-Specific Neural Reorganization and Predicts Motor...

Area of Science:

• Neuroscience
• Rehabilitation Medicine
• Biomedical Engineering

Background:

• Stroke frequently causes motor impairments due to disrupted cortical connectivity, posing a significant neurorehabilitation challenge.
• Understanding brain network reorganization is crucial for optimizing motor recovery strategies post-stroke.

Purpose of the Study:

• To evaluate brain network reorganization following a 10-week gait rehabilitation program in stroke survivors.
• To investigate the relationship between network changes and motor recovery.
• To assess the efficacy of Exoskeleton-Assisted Rehabilitation (ER) versus Standard of Care (SOC).

Main Methods:

• Resting-state electroencephalography (EEG) recorded from 22 stroke survivors and 22 healthy controls pre- and post-intervention.
• Directed functional connectivity estimated using Partial Directed Coherence (PDC), followed by graph theory and laterality index (LI) analysis.
• Machine learning classification applied to frequency-specific EEG features for group differentiation.

Main Results:

• Stroke patients showed altered baseline connectivity in motor and premotor areas compared to controls.
• Post-intervention, increased connectivity was observed in motor-related areas, correlating with Fugl-Meyer score improvements (r=0.64, p=0.001).
• Machine learning models achieved high accuracy (93.18%) in differentiating groups using alpha and gamma band EEG features.

Conclusions:

• Source-level EEG connectivity and frequency-specific features serve as sensitive biomarkers for neuroplasticity.
• The rehabilitation program enhanced motor network integration and hemispheric balance, correlating with clinical gains.
• EEG-based machine learning offers scalable solutions for precise neurorehabilitation assessment.