Machine Learning Classification to Identify the Stage of Brain-Computer Interface Therapy for Stroke Rehabilitation...

Area of Science:

• Neuroscience
• Rehabilitation Medicine
• Biomedical Engineering

Background:

• Stroke survivors often experience persistent motor impairments.
• Brain-computer interface (BCI) therapy shows promise for motor recovery.
• The effects of BCI on non-motor brain networks post-stroke are not well understood.

Purpose of the Study:

• To investigate changes in resting-state functional connectivity (rs-FC) in stroke patients undergoing BCI therapy.
• To identify functional network changes outside the motor system.
• To use machine learning to classify participants by therapy stage based on rs-FC.

Main Methods:

• Twenty chronic stroke participants with upper-extremity impairment received BCI therapy.
• Resting-state functional MRI (rs-fMRI) data were collected pre- and post-intervention.
• A support vector machine (SVM) classifier analyzed rs-FC from 236 brain seeds.

Main Results:

• The SVM classifier achieved 92.5% cross-validation accuracy in distinguishing therapy stages.
• Non-motor networks (fronto-parietal, default mode, subcortical, visual) significantly contributed to classification.
• More functional changes strengthened than weakened from pre- to post-therapy, affecting both motor and non-motor regions.

Conclusions:

• BCI therapy influences functional connectivity beyond motor networks in stroke survivors.
• These findings support BCI's potential for broader stroke rehabilitation.
• Understanding network changes can optimize future BCI intervention designs.