ARTIFICIAL INTELLIGENCE IN POSTURE ASSESSMENT: A SYSTEMATIC REVIEW

Abstract

Background: In musculoskeletal health, posture is crucial because it affects biomechanics, pain, and functional performance. Because of sedentary lifestyles and extended screen time, poor posture, including kyphosis, lordosis, and forward head posture, is becoming more common. In posture evaluation, artificial intelligence (AI), especially machine learning (ML) and computer vision, has become a revolutionary tool. AI makes it possible to assess human posture objectively, automatically, and in real time utilizing depth cameras, wearable sensors, and image systems. Methods: A systematic search was conducted across PubMed, Scopus, Google Scholar, and ScienceDirect following PRISMA guidelines. Studies involving AI-driven posture assessment tools (e.g., APECS, MediaPipe, OpenPose) were included. Eligible studies assessed reliability, validity, or clinical applicability compared with traditional techniques. Results: AI-based systems demonstrated strong reliability and validity, with several studies reporting high intraclass correlation coefficients (ICC > 0.90) and significant correlations with gold-standard methods (r = 0.71–0.99). These systems provide objective measurements through automated landmark detection and angle calculations. Additional advantages include reduced assessment time, improved consistency, automated reporting, and applicability in tele rehabilitation. Conclusion: Traditional posture assessment methods in physiotherapy rely heavily on visual observation, goniometry, and manual palpation, which can be subjective and inconsistent. AI systems improve accuracy, reliability, and accessibility of assessment, especially in tele-rehabilitation settings.