ARTIFICIAL INTELLIGENCE AND RADIOMICS IN JOINT AND SPINE PAIN IMAGING: CURRENT CLINICAL APPLICATIONS, QUANTITATIVE BIOMARKERS, AND FUTURE PERSPECTIVES

Abstract

Background: Joint and spine pain are major contributors to global disability and healthcare burden. Although conventional imaging modalities, including radiography, computed tomography (CT), magnetic resonance imaging (MRI), and ultrasound, remain central to diagnosis, they rely predominantly on qualitative assessment and often demonstrate limited correlation with clinical symptoms. Artificial intelligence (AI) and radiomics have emerged as transformative technologies that enable automated image analysis and extraction of quantitative imaging biomarkers. Objective: This review summarizes the current applications of AI and radiomics in joint and spine pain imaging, highlighting quantitative biomarkers, clinical applications, evidence from recent studies, challenges, and future directions. Methods: A comprehensive narrative review of recent literature was performed, focusing on machine learning, deep learning, radiomics, and quantitative imaging biomarkers in musculoskeletal imaging. Studies involving degenerative, inflammatory, and traumatic disorders of the joints and spine were critically evaluated. Results: AI has demonstrated high diagnostic performance in automated detection, disease classification, severity grading, and outcome prediction for conditions such as osteoarthritis, degenerative disc disease, lumbar spinal stenosis, sacroiliitis, meniscal injuries, and rotator cuff disorders. Radiomics provides objective characterization of cartilage, bone, intervertebral discs, bone marrow, muscles, and synovium, facilitating early disease detection, risk stratification, and personalized treatment planning. However, widespread clinical implementation is limited by small datasets, variability in imaging protocols, lack of standardization, segmentation challenges, limited external validation, reproducibility concerns, and regulatory issues. Conclusion: AI and radiomics are transforming musculoskeletal imaging by enabling quantitative, data-driven assessment and precision medicine. Future advances in multicenter validation, standardized radiomics workflows, explainable AI, federated learning, and integration of imaging biomarkers with clinical and molecular data are expected to accelerate clinical translation and improve diagnosis, prognostication, and individualized management of patients with joint and spine pain.