From pixels to prediction: improving imaging through AI in rheumatology

Imaging plays a central role in the diagnosis, monitoring, and prediction of rheumatic and musculoskeletal diseases (RMDs), yet its clinical and regulatory impact remains limited by variability, the lack of standardisation and metrics for quantification, and slow clinical translation. Artificial intelligence (AI) offers unique opportunities to overcome these barriers through automated segmentation, quantitative feature extraction, and multimodal data integration. The AutoPiX consortium seeks to design and validate machine-learning-driven imaging pipelines to expedite the translation of imaging data into quantitative biomarkers, facilitate their systematic characterisation, and promote their use in enhancing patient stratification and bringing imaging-based endpoints closer to regulatory endorsement in clinical trials. This paper outlines unmet needs, the potential translational impact of AI in RMDs and the perspectives of key stakeholders, including clinicians, patients, regulators, and industry partners. We call for harmonised imaging protocols, large-scale imaging datasets, cross-disciplinary collaboration, and early regulatory engagement to realise the transformative potential of AI-enhanced imaging biomarkers.