EEG Ensemble Learning Framework for Prognostic Prediction of Post-rTMS Motor Recovery in Stroke

Area of Science:

• Neuroscience
• Rehabilitation Medicine
• Machine Learning

Background:

• Repetitive transcranial magnetic stimulation (rTMS) shows promise for post-stroke rehabilitation by modulating brain networks.
• Inter-individual variability in patient response to rTMS, linked to neural reorganization, hinders clinical effectiveness.
• Predicting motor recovery outcomes is crucial for tailoring rTMS interventions.

Purpose of the Study:

• To develop and evaluate an EEG ensemble learning framework (EEG-ELF) for predicting motor recovery in stroke patients receiving rTMS.
• To assess the framework's ability to correlate pre-treatment EEG features with actual motor function improvement.
• To enable data-driven patient stratification for personalized rTMS therapy.

Main Methods:

• Utilized electroencephalography (EEG) recordings from 20 subcortical stroke patients during a motor imagery task, collected pre- and post-rTMS intervention.
• Applied an EEG ensemble learning framework (EEG-ELF) for outcome prediction.
• Evaluated performance using regression and classification with leave-one-subject-out cross-validation against upper-limb Fugl-Meyer Assessment (FMA) scores.

Main Results:

• EEG-ELF demonstrated a strong correlation between predicted and actual motor recovery rates (Pearson's R = 0.92, p < 0.001).
• The framework achieved 95% accuracy in classifying patients into good and poor prognosis groups.
• Pre-treatment EEG features were found to be significant predictors of individual motor outcomes.

Conclusions:

• EEG-ELF accurately predicts individual motor recovery outcomes in stroke patients undergoing rTMS.
• The framework's predictive capability can guide personalized rTMS interventions.
• Data-driven patient stratification using EEG-ELF holds potential for optimizing stroke rehabilitation strategies.