Artificial Intelligence and Wearable Sensor-Based Prediction of Upper-Limb Overuse Injuries in Wheelchair Para Athletes: Implications for Sports Physiotherapy

Abstract

Upper-limb overuse pathology, particularly of the shoulder and wrist, is the leading cause of pain and impaired performance among wheelchair para athletes, arising from the paradox of using load-bearing joints originally suited for mobility rather than sustained propulsive force. Traditional clinical screening relies on intermittent, subjective assessment and laboratory-based biomechanical testing, both of which are poorly suited to capturing the cumulative, ecologically valid loading histories that precipitate overuse injury. Recent convergence of low-cost inertial measurement units (IMU), surface electromyography (EMG), and pushrim force sensors with machine learning (ML) and deep learning (DL) analytics has opened a pathway toward continuous, field-based injury risk surveillance. This review synthesises current evidence on artificial intelligence (AI)-based movement analysis and wearable sensor systems for predicting upper-limb overuse injury in wheelchair-dependent para athletes. We examine the epidemiological and biomechanical rationale for shoulder and wrist vulnerability, the sensor modalities and AI architectures reported in the literature, and the reported predictive performance of these systems, before discussing the translational implications for sports physiotherapy practice, including workload monitoring, technique correction, and early referral pathways. The review identifies persistent gaps — a scarcity of studies conducted specifically in wheelchair-sport (rather than able-bodied or activities-of-daily-living) populations, limited external validation, heterogeneous outcome definitions, and minimal integration into physiotherapist-facing clinical workflows — and proposes a research and implementation agenda to close them.

Introduction

The text reviews the use of wearable sensors and artificial intelligence (AI) to monitor and predict upper-limb overuse injuries in wheelchair para-athletes.

• Problem: Wheelchair athletes rely heavily on their shoulders, elbows, and wrists for both sports and daily activities. Repetitive propulsion and high-force movements therefore create a high risk of overuse injuries such as shoulder impingement, rotator cuff tendinopathy, and carpal tunnel syndrome.
• Limitations of current assessment: Traditional methods, including physiotherapist screening, pain questionnaires, and laboratory motion capture, are periodic, expensive, and unable to continuously measure the cumulative workload responsible for injury.
• Wearable sensors: Technologies such as IMUs, surface EMG, and instrumented pushrims/force sensors can continuously capture movement, muscle activity, and propulsion forces during real-world wheelchair use.
• AI and machine learning: Algorithms including Random Forest, XGBoost, SVM, CNN, LSTM, and Bi-LSTM can analyse sensor data to identify movement patterns and estimate shoulder loading. Some studies have reported classification accuracies above 90–95%.
• Important biomechanical factors: Higher stroke frequency, shorter push angles, greater peak forces, force rate-of-rise, and increased movement demands during sprinting and turning are associated with greater shoulder loading and pain.
• Research gap: Although sensor-based activity and load classification is highly promising, there is still limited research directly predicting clinically diagnosed overuse injuries in competitive wheelchair athletes. Many existing studies focus on able-bodied participants, activities of daily living, or surrogate measures such as pain and shoulder load.
• Role of physiotherapy: AI-based wearable monitoring could provide physiotherapists with objective, continuous information about workload, propulsion technique, muscle effort, and injury risk, enabling earlier intervention and more personalised injury-prevention strategies.

Conclusion

Wearable sensor and AI/ML approaches have demonstrated strong technical feasibility for classifying wheelchair-related upper-limb activity and estimating joint load, and an adjacent, more mature body of sports-injury-prediction literature reports high classification accuracy across other athlete populations. However, direct, prospectively validated prediction of upper-limb overuse injury specifically within wheelchair para-athlete cohorts remains an emerging rather than established evidence base. For sports physiotherapists, current systems are best regarded as objective monitoring adjuncts to, rather than replacements for, clinical assessment, with the greatest near-term value in individualised workload management and early technique-based risk flagging.

References

[1] I. P. Salzmann, T. Rietveld, R. Togni, S. J. Briley, V. L. Goosey-Tolfrey, and W. H. K. de Vries, \"Exploring the biomechanical link between wheelchair propulsion, shoulder injury and shoulder pain: A scoping review,\" J. Biomech., vol. 185, art. 112678, May 2025. [2] \"Shoulder complaints in wheelchair athletes: A systematic review,\" PLOS ONE, 2017. [3] L. H. V. van der Woude, H. E. J. Veeger, and A. J. Dallmeijer, \"Wheelchair propulsion biomechanics: implications for wheelchair sports,\" 2001. [4] B. J. H. Beirens, F. M. Bossuyt, U. Arnet, L. H. V. van der Woude, and W. H. K. de Vries, \"Alterations in shoulder kinematics are associated with shoulder pain during wheelchair propulsion sprints,\" 2022. [5] B. Mason et al., \"Constraints influencing sports wheelchair propulsion performance and injury risk,\" BMC Sports Sci. Med. Rehabil., vol. 5, art. 3, 2013. [6] Muscle forces analysis in the shoulder mechanism during wheelchair propulsion, National Cheng Kung Univ. / China Medical Univ., 2004. [7] \"Alterations in shoulder kinematics are associated with shoulder pain during wheelchair propulsion sprints,\" PMC, 2022. [8] B. J. H. Beirens et al., \"Shoulder pain is associated with rate of rise and jerk of the applied pushrim force,\" 2022. [9] \"Machine Learning-Based Classification of Wheelchair Task Intensity for Injury Risk Prediction,\" Automation, 2026. [10] M. Morrow et al., \"Estimation of Manual Wheelchair-Based Activities in the Free-Living Environment using a Neural Network Model with Inertial Body-Worn Sensors,\" 2022. [11] \"Classification of Wheelchair Related Shoulder Loading Activities from Wearable Sensor Data: A Machine Learning Approach,\" Sensors, vol. 22, art. 7404, 2022. [12] S. Amrein, C. Werner, U. Arnet, and W. H. K. de Vries, \"Machine-Learning-Based Methodology for Estimation of Shoulder Load in Wheelchair-Related Activities Using Wearables,\" Sensors, vol. 23, art. 1577, 2023. [13] \"Diagnostic Applications of AI in Sports: A Comprehensive Review of Injury Risk Prediction Methods,\" Diagnostics, vol. 14, no. 22, art. 2584, 2024. [14] Z. Xu, W. Sun, H. Qian, and M. Yao, \"Construction and application of a model for predicting athletes\' injury risk based on machine learning,\" BMC Med. Inform. Decis. Mak., 2025. [15] \"Time-to-Injury Forecasting in Elite Female Football: A DeepHit Survival Approach,\" arXiv, 2026. [16] \"Artificial intelligence and wearable sensors in sports injury risk prediction: current status and future perspectives,\" PubMed, 2025. [17] \"Artificial intelligence-integrated wearable technology in sports medicine injury prevention and rehabilitation: A narrative review,\" J. Sport Rehabil. Sci., 2026. [18] \"Multi-modal fusion of medical imaging and biomechanical data using attention-based Swin-UNet and LSTM for sports injury prediction,\" PMC, 2025. [19] \"Application of Artificial Intelligence for Predicting Sports Injuries and Customizing Personalized Prevention Strategies: A Scoping Review,\" PubMed, 2026. [20] \"Machine learning approaches to injury risk prediction in sport: a scoping review with evidence synthesis,\" Br. J. Sports Med., 2024. [21] \"Sensor-enhanced wearables and automated analytics for injury prevention in sports,\" ScienceDirect, 2024.

Copyright

Copyright © 2026 Dr. Ruby. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.