Engagement Enhancement Based on Human-in-the-Loop Optimization for Neural Rehabilitation

Area of Science:

• Neuroscience
• Rehabilitation Engineering
• Human-Computer Interaction

Background:

• Individual physiological and neurological differences create challenges in enhancing patient engagement for neural rehabilitation.
• Task difficulty significantly impacts engagement, but optimizing it dynamically based on individual responses is complex.
• Existing methods struggle to adapt training tasks to evolving patient responses during rehabilitation.

Purpose of the Study:

• To propose and validate a human-in-the-loop optimization method for enhancing patient engagement in neural rehabilitation.
• To personalize training task difficulty to maximize patient engagement and improve rehabilitation outcomes.
• To simultaneously evaluate physical and physiological responses, including muscle activation and neural activity.

Main Methods:

• An interactive speed-tracking riding game was developed with parameterized reference speed curves (RSCs) to adjust task difficulty.
• An objective function integrated tracking accuracy and surface electromyogram (sEMG)-based muscle activation for response evaluation.
• A covariance matrix adaptation evolution strategy was employed for periodic optimization of RSC parameters.

Main Results:

• The proposed human-in-the-loop optimization method successfully enhanced patient engagement.
• Significant improvements in both surface electromyogram (sEMG)-based motor engagement and electroencephalography (EEG)-based neural engagement were observed.
• The optimized training tasks maintained a high level of engagement throughout the rehabilitation process.

Conclusions:

• Personalized training task optimization using human-in-the-loop strategies is effective for neural rehabilitation.
• The method demonstrates the potential to significantly improve both motor and neural engagement in patients.
• This approach offers a promising direction for developing more adaptive and effective rehabilitation technologies.