Quantifying User Engagement During an Upper Limb Rehabilitation Task

Area of Science:

• Robotics
• Human-Computer Interaction
• Rehabilitation Engineering

Background:

• Patient engagement is crucial for effective post-stroke robotic rehabilitation.
• Limited research exists on modulating and quantifying engagement during therapy.

Purpose of the Study:

• To develop and evaluate a virtual reality (VR)-integrated robot-assisted system for upper limb rehabilitation.
• To enable simultaneous modulation and monitoring of user engagement.
• To investigate the effectiveness of physiological and behavioral indicators for engagement estimation.

Main Methods:

• A VR-integrated robot-assisted system was used for a line tracing task.
• Task difficulty was modulated via shape complexity and force noise.
• Engagement was estimated using physiological (GSR, pupil diameter) and behavioral (eye blink, gaze) signals.
• A Game Engagement Questionnaire (GEQ) was used for benchmarking.
• Twenty healthy subjects participated.

Main Results:

• Behavioral signals were more informative for predicting engagement than physiological signals.
• An optimal 11-second analysis window was identified for accurate engagement metrics (MAE = 0.73, r = 0.42).
• Peak engagement, aligning with flow theory, occurred when task difficulty matched user skill (Gaussian model: R2 = 0.76, RMSE = 0.18).

Conclusions:

• Behavioral measurements offer a reliable, non-invasive method for estimating engagement during rehabilitation tasks.
• This approach supports the development of adaptive systems that adjust difficulty to optimize patient engagement.
• Findings pave the way for enhanced robotic rehabilitation strategies.