Predicting Motor Recovery in the Subacute Phase of Stroke: The Role of Passive Motor Task and Electroencephalogram Brain Networks

OBJECTIVE: In stroke rehabilitation, the identification of reliable prognostic factors remains a challenge for effective rehabilitation strategies. This study investigates which motor task has the most predictive utility based on electroencephalogram (EEG) characteristics during the subacute phase of stroke. METHODS: Data were collected from 10 subacute stroke patients performed different motor tasks (passive, robotic movement; active, voluntary execution; motor imagery [MI]) and a nonmotor task (No-Go) at two time points: T1, approximately 3 weeks after stroke onset, and T2, 2 months after T1. EEG power and network characteristics were extracted for each task, and prognostic models were constructed with nested leave-one-subject-out (LOSO) cross-validation (CV) to avoid information leakage. The passive task also showed less overactivation and hyperconnectivity in the early subacute phase relative to the active and MI tasks. This performance advantage was replicated across six analytic procedures (2 feature-selection × 3 regression strategies), supporting the robustness of the task comparison within the present dataset. CONCLUSIONS: These findings suggest that EEG features obtained during a robot-assisted passive task may provide exploratory prognostic utility for motor recovery prediction in the early subacute phase of stroke.Impact StatementThe passive task may offer a practical approach for assessing neurophysiological markers of recovery potential, supporting the development of more personalized rehabilitation strategies and earlier prognostic evaluation in stroke patients.