Design and Validation of Vision-Based Exercise Biofeedback for Tele-Rehabilitation
Ali Barzegar Khanghah1,2, Geoff Fernie1,2,3, Atena Roshan Fekr1,2
1KITE Research Institute, Toronto Rehabilitation Institute, University Health Network, 550 University Ave, Toronto, ON M5G 2A2, Canada.
Sensors (Basel, Switzerland)
|February 11, 2023
Summary
This study introduces a vision-based biofeedback system for tele-rehabilitation, using 3D Convolution Neural Networks (3D-CNN) to automatically assess exercise quality. The system accurately identifies correct and incorrect movements, enhancing remote patient care.
Area of Science:
- Rehabilitation Technology
- Computer Vision
- Artificial Intelligence
Background:
- Tele-rehabilitation offers accessible and cost-effective patient monitoring from home.
- Automated exercise guidance in tele-rehab platforms can significantly improve outcomes.
- A novel vision-based biofeedback system is needed to objectively assess exercise quality.
Purpose of the Study:
- To design and validate a vision-based biofeedback system for identifying the quality of rehabilitation exercises.
- To enable patients to refine movements for optimal care through automated feedback.
- To leverage artificial intelligence for enhanced tele-rehabilitation services.
Main Methods:
- Utilized a pre-trained 3D Convolution Neural Network (3D-CNN) for video analysis.
- Employed an open dataset of 30 participants performing nine distinct exercises.
- Exercises were meticulously labeled as 'Correctly' or 'Incorrectly' executed by five clinicians.
Main Results:
- Achieved high average accuracy: 90.57% (10-Fold) and 83.78% (LOSO) cross-validation.
- Obtained average F1-scores of 71.78% (10-Fold) and 60.64% (LOSO) cross-validation.
- The 3D-CNN system demonstrated robust classification of rehabilitation exercise quality.
Conclusions:
- The developed vision-based biofeedback system effectively classifies rehabilitation exercise quality.
- This technology can provide crucial feedback to patients, aiding in movement pattern modification.
- The system holds significant potential for improving tele-rehabilitation outcomes and patient engagement.
