Exploratory development of human-machine interaction strategies for post-stroke upper-limb rehabilitation

Area of Science:

• Biomedical Engineering
• Rehabilitation Robotics
• Neuroscience

Background:

• Stroke significantly impacts upper limb function, necessitating advanced rehabilitation solutions.
• Human-machine interaction (HMI) is crucial for effective post-stroke motor recovery.
• Current active HMI strategies face challenges in distinguishing subtle motion intentions from involuntary movements.

Purpose of the Study:

• To develop and validate an upper-limb exoskeleton system (UarDus) with novel HMI strategies for stroke rehabilitation.
• To enable personalized and real-time active rehabilitation by accurately capturing patient's motion intentions.
• To provide both passive and active rehabilitation options within a single integrated system.

Main Methods:

• Designed a 14-degrees of freedom (DoFs) upper-limb exoskeleton based on human physiological structure and scapulohumeral rhythm.
• Implemented three HMI strategies: robot-in-charge, therapist-in-charge, and patient-in-charge.
• Developed a deep learning model (CNN-Transformer) for subtle motion intention recognition and used Discrete Wavelet Transform (DWT) for safety monitoring.

Main Results:

• The exoskeleton demonstrated a comfortable dynamic wear experience with well-matched motion curves.
• Passive strategies showed good real-time control, while the active strategy achieved 99.7% classification accuracy for 15 actions.
• The system effectively identified subtle motion intentions and ensured safety through DWT-based surge detection.

Conclusions:

• The UarDus exoskeleton system with integrated passive and active HMI modalities offers a viable solution for post-stroke upper limb rehabilitation.
• Proof-of-concept study validated the system's capability for safe, personalized, real-time rehabilitation training.
• The system provides a dynamically comfortable wear experience, enhancing patient engagement and recovery potential.