Development of an artificial intelligence based diagnostic model to detect and classify osteoarthritis on plain X-ray

Knee osteoarthritis (OA) represents a major cause of morbidity and functional impairment, particularly among older adults. Globally, the prevalence of knee OA is estimated at approximately 16%, with an incidence of 203 per 10,000 person-years, increasing markedly with age. In India, nearly 45% of individuals above 65 years report symptoms, while up to 70% show radiological evidence of disease. Despite its high burden, early diagnosis of knee OA remains challenging. Plain radiography is the most widely used diagnostic modality; however, interpretation is largely subjective and dependent on the clinician’s experience, leading to poor inter-observer reliability and underdiagnosis, especially in early-stage disease. It is estimated that nearly 15% of early knee OA cases remain undetected in routine clinical practice.

To address these challenges, the proposed study aims to develop and evaluate an artificial intelligence (AI)–driven diagnostic model for the detection and classification of knee osteoarthritis using plain anteroposterior knee X-ray images. The study is designed as a superiority clinical trial utilizing an AI-based diagnostic tool, with a planned sample size of 1,600 X-ray images over a duration of two years. Adult patients (>18 years) presenting to the Orthopaedics Department of AIIMS Deoghar with chronic knee pain of more than three months and preserved knee range of motion will be included, while those with recent lower-extremity surgery will be excluded.

The proposed AI model has the potential to enhance diagnostic accuracy, reduce observer variability, and facilitate early and precise identification of knee OA. Early detection may enable timely interventions that slow disease progression, reduce pain, and improve functional outcomes. Importantly, in the Indian context—where approximately 70% of the population resides in rural or remote areas and there is a shortage of specialized healthcare professionals—such a tool could significantly support clinical decision-making and reduce the burden on healthcare providers. Additionally, the project aligns with national priorities such as the Digital Health Mission by promoting cost-effective, accessible, and equitable healthcare solutions. Overall, this AI-driven approach holds promise for democratizing musculoskeletal diagnostics and improving knee OA management across diverse healthcare settings.