A Narrative Review on AI-assisted Clinical Decision Support System in Physiotherapy Practice: Computer Vision-based Posture Assessment and its Clinical Validation

1. Associate Professor, Department of Physiotherapy, MM Institute of Physiotherapy and Rehabilitation, Maharishi Markandeshwar (Deemed to be University), Mullana, Ambala, Haryana, India. 2. Associate Professor, Department of Physiotherapy, MM Institute of Physiotherapy and Rehabilitation, Maharishi Markandeshwar (Deemed to be University), Mullana, Ambala, Haryana, India. 3. Director and Chief Physiotherapist, Department of Physiotherapy, Relief Care Physiotherapy Centre, Burari, Delhi, India. 4. Assistant Professor, College of Physiotherapy, Baba Mastnath University, Rohtak, Haryana, India.

Abstract

Artificial Intelligence (AI)-assisted Computer Vision (CV) systems are applied in physiotherapy for objective, markerless assessment of posture and movement. The conventional clinical evaluation relies on visual inspection and manual goniometry; therefore, it is subject to inter-rater variability and measurement discrepancy. The CV frameworks employ Red, Green, Blue (RGB) cameras, depth sensors, and hybrid optical-inertial configurations to detect skeletal landmarks and calculate joint angles, range of motion, symmetry indices, and temporal kinematic variables. The pose estimation models such as MediaPipe, OpenPose, and YOLO-Pose permit real-time landmark detection. The Machine Learning (ML) and Deep Learning (DL) algorithms classify movement quality and generate structured feedback for rehabilitation sessions. The incorporation of CV into physiotherapy therefore, provides quantitative assessment of posture and movement that corresponds with conventional clinical measurement methods and supports systematic rehabilitation monitoring. The review examines the technological architecture, validation standards, and clinical applicability of AI-assisted CV systems in physiotherapy practice and evaluates their role in clinical decision support and movement assessment within structured rehabilitation programmes.

Artificial intelligence, Biomechanical analysis, Electromyography, Forward head posture, Goniometry, Markerless motion capture