Clinical Report: Classification of Rehabilitation Movement Quality Using Skeleton Data

Overview

This study establishes a benchmark for assessing rehabilitation movement quality using skeleton data, addressing issues of subject leakage and class imbalance. The multilayer perceptron achieved the highest accuracy of 0.869.

Background

Automated assessment of rehabilitation movements is crucial for scalable physiotherapy monitoring, particularly as traditional methods can be resource-intensive and inconsistent. The ability to objectively evaluate movement quality using three-dimensional skeleton recordings can enhance monitoring. However, existing methodologies often suffer from biases due to subject leakage and imbalanced datasets, necessitating a rigorous benchmarking approach.

Data Highlights

Key Findings

• The study utilized a leakage-controlled, subject-independent benchmark on the IntelliRehabDS dataset.
• A total of 2,575 sequences from 29 participants were analyzed, with 2,047 correct and 528 incorrect executions across nine gestures.
• The multilayer perceptron model achieved the highest accuracy of 0.869.
• Head-motion and bilateral-asymmetry features were significant in distinguishing between correct and incorrect executions.
• ST-SkelNet did not consistently outperform classical models despite its advanced architecture.

Clinical Implications

The findings highlight the importance of using leakage-controlled methodologies in evaluating rehabilitation movement quality.

Conclusion

This benchmark study provides a reproducible framework for assessing rehabilitation movement quality, emphasizing the need for rigorous evaluation protocols in future research.

Related Resources & Content

1. npj Digital Medicine, 2025 -- Markerless 3D Pose Analysis for Scalable Remote Assessment of Gait Kinematics
2. DIGITAL HEALTH, 2025 -- Toward intelligent rehabilitation: Multimodal human pose modeling with parametric meshes and graph-based temporal reasoning
3. Deep Learning-Based Segmentation for Automated Detection of Tibial Implant Loosening, 2025
4. npj Digital Medicine — Affordable AI-Powered Exergame for Stroke Rehabilitation and Upper-Limb Function Evaluation
5. Marketing Submission Recommendations for a Predetermined Change Control Plan for Artificial Intelligence-Enabled Device Software Functions | FDA
6. 3 Early-use HealthTech guidance assessments | NICE HealthTech programme manual | Guidance | NICE
7. The ATA’s Principles for Delivering Telerehabilitation Services - REVISED - ATA
8. Reliability and validity of current computer vision based motion capture systems in gait analysis: A systematic review - PubMed
9. Reliability and validity of lower extremity and trunk kinematics measured with markerless motion capture during sports-related and functional tasks: A systematic review
10. Frontiers | Validity and reliability of single camera markerless motion capture systems with RGB-D sensors for measuring shoulder range-of-motion: a systematic review
11. Validation of a markerless motion capture app for automated scoring of sit-to-stand, timed up and go, and short physical performance battery tests in adults with chronic disease | PLOS Digital Health
12. JMIR mHealth and uHealth - Effects of a Computer Vision–Based Exercise Application for People With Knee Osteoarthritis: Randomized Controlled Trial
13. 3D pose estimation for scalable remote gait kinematics assessment