Automated Control of Rehabilitation Process in Physical Therapy Using a Novel Human Skeleton-Based Balanced Time...

Area of Science:

• Biomedical Engineering
• Rehabilitation Science
• Computer Vision

Background:

• Physical therapy is crucial for recovery from various medical conditions.
• Accurate monitoring of patient exercise execution and progress is essential for effective rehabilitation.
• Current methods for assessing physical therapy quality can be subjective and labor-intensive.

Purpose of the Study:

• To develop and evaluate a novel automated system for real-time quality assessment of physical therapy exercises.
• To quantitatively measure the similarity between patient and physiotherapist exercise performance.
• To provide objective feedback for personalized rehabilitation and progress tracking.

Main Methods:

• Utilized computer vision and a Human Skeleton-based Balanced Time Warping algorithm to analyze skeletal motion data from RGB-D videos.
• Computed pose dissimilarity using a motion-dependent, weighted Euclidean distance between 3D skeletal models.
• Combined posture and dynamics dissimilarity components and validated against a custom dataset of 136 recordings.

Main Results:

• The system demonstrated strong clustering of exercise accuracy levels (good, intermediate, bad).
• A high Spearman's rank correlation coefficient (0.977) was found between computed dissimilarity and actual execution accuracy.
• The proposed dissimilarity measure effectively orders patient performance relative to a reference physiotherapist.

Conclusions:

• The developed system provides an objective, accurate, and scalable method for evaluating physical therapy exercise quality.
• This technology can enhance rehabilitation by offering real-time feedback, reducing therapist workload, and enabling remote monitoring.
• Future clinical validation with specific patient groups (e.g., post-stroke, cardiac) is planned.