Robot-Assisted Mirror Rehabilitation for Post-Stroke Upper Limbs: A Personalized Control Strategy

Area of Science:

• Rehabilitation Engineering
• Robotics in Medicine
• Neurorehabilitation

Background:

• Traditional mirror therapy for stroke rehabilitation has limitations, including rigid movement mapping and lack of personalization.
• Existing robotic systems may not fully integrate multimodal biofeedback for adaptive therapy.

Purpose of the Study:

• To propose a robot-assisted mirror rehabilitation framework that overcomes the limitations of traditional mirror therapy.
• To enhance personalization and movement mapping in stroke rehabilitation through multimodal biofeedback integration.

Main Methods:

• Developed a dual-modal feature fusion network using cross-attention to integrate kinematic data and surface electromyography (sEMG) signals.
• Incorporated a dynamically generated mirror ratio coefficient (λ) into an exoskeleton's admittance control loop.
• Utilized dynamic movement primitives (DMPs) for trajectory generation and a compliant control strategy with dynamic constraints for personalized and safe rehabilitation.

Main Results:

• The proposed system demonstrated superior performance in bilateral trajectory covariance metrics compared to traditional mirror therapy.
• Significant improvements were observed in the mirror symmetry index and muscle activation levels.
• The framework achieved personalized rehabilitation trajectory planning and safe human-robot interaction.

Conclusions:

• The robot-assisted mirror rehabilitation framework with multimodal biofeedback offers a more personalized and effective approach to stroke rehabilitation.
• Integration of kinematic and sEMG data via a cross-attention network enhances the adaptiveness and efficacy of mirror therapy.
• This technology shows promise for improving motor recovery and functional outcomes in stroke survivors.