A Novel Feedback-Based Compensation Reduction With Upper Body Reconstruction for Upper-Limb Rehabilitation.

Area of Science:

• Rehabilitation Engineering
• Neurorehabilitation
• Biomedical Engineering

Background:

• Compensatory movements hinder motor recovery in stroke patients undergoing upper-limb rehabilitation.
• Existing vision-based systems have limitations in tracking accuracy and integrating trunk and arm movements.

Purpose of the Study:

• To develop and validate a real-time system for detecting and reducing compensatory movements during robot-assisted upper-limb rehabilitation.
• To improve motor recovery and movement quality in stroke patients by providing integrated trunk and arm movement feedback.

Main Methods:

• Developed a sensor-fusion system for upper-body reconstruction and a machine learning classifier to detect compensatory movements.
• Integrated the system with an upper-limb rehabilitation robot to provide real-time audio-visual feedback.
• Validated the system with 18 stroke patients in experimental and control groups during reaching tasks.

Main Results:

• The machine learning classifier achieved high performance (F1-scores of 0.85 and 0.77).
• The experimental group significantly reduced compensatory movements (58.9% to 38.2%, p=0.016).
• Improved trajectory accuracy (p=0.0052) and reduced trunk movement magnitude (p<0.05) were observed in the experimental group.

Conclusions:

• The developed system effectively reduces compensatory movements in stroke patients during upper-limb rehabilitation.
• The system enhances movement quality and shows significant potential for clinical application in stroke recovery.
• Integrated sensor-fusion and machine learning offer a promising approach for personalized rehabilitation feedback.