An Investigation of Artificial Intelligence for Movement Quality Assessment in Physical Rehabilitation

The quality of physical rehabilitation training is crucial for patients’ recovery outcomes. However, traditional rehabilitation methods have significant limitations in objective and quantitative assessment. To address this issue, this paper aims to conduct a systematic review of the application of artificial intelligence (AI) in the assessment of rehabilitation movement quality. This paper systematically reviews the technological evolution path from traditional machine learning methods that rely on manual feature engineering to deep learning models that can automatically learn complex spatiotemporal features from raw data. The review finds that although deep learning methods show great potential in assessment accuracy, they still face major challenges such as poor model interpretability, insufficient generalization ability in real-world environments, and difficulties in aligning with clinical assessment criteria. Future development directions should focus on expert systems that integrate domain knowledge, the construction of large-scale standardized datasets, and the development of multi-modal fusion models for clinical objectives. This review provides researchers and clinicians in this field with perspectives on the current state of technology, core challenges, and future paths, in order to promote the development of a new paradigm of personalized, data-driven precision rehabilitation.