Artificial intelligence in axial spondyloarthritis imaging: a review of current evidence.
Authors
Affiliations (5)
Affiliations (5)
- Department of Diagnostic and Interventional Radiology, School of Medicine and Health, Technical University of Munich, Munich, Germany.
- Department of Gastroenterology, Infectious Diseases and Rheumatology, Charité - Universitätsmedizin Berlin, corporate member of Freie Universität Berlin and Humboldt-Universität zu Berlin, Berlin, Germany.
- Division of Rheumatology, Department of Medicine, University of Toronto, Toronto, Ontario, Canada; Schroeder Arthritis Institute, University Health Network, Ontario, Toronto, Canada.
- Department of Diagnostic and Interventional Radiology, School of Medicine and Health, Technical University of Munich, Munich, Germany; National Center for Tumor Diseases West, Essen, Germany; Institute of Artificial Intelligence in Medicine, University Hospital Essen, Essen, Germany; Department of Diagnostic and Interventional Radiology and Neuroradiology, University Hospital Essen, Essen, Germany.
- Department of Gastroenterology, Infectious Diseases and Rheumatology, Charité - Universitätsmedizin Berlin, corporate member of Freie Universität Berlin and Humboldt-Universität zu Berlin, Berlin, Germany; Division of Rheumatology, Department of Medicine, University of Toronto, Toronto, Ontario, Canada; Schroeder Arthritis Institute, University Health Network, Ontario, Toronto, Canada. Electronic address: [email protected].
Abstract
The diagnosis of axial spondyloarthritis (axSpA) remains challenging and is frequently delayed. Imaging is central to diagnosis, but relevant inflammatory and structural abnormalities may be subtle, heterogeneous, or overlap with findings related to mechanical stress, anatomical variation, and degeneration. Artificial intelligence (AI), particularly deep learning, is therefore being investigated to support image interpretation, lesion detection, quantification, and standardized assessment. We synthesize evidence from 12 studies applying convolutional neural networks and related architectures to magnetic resonance imaging and radiographs, automate imaging scores, and combine imaging-derived information with clinical variables. Several models achieved performance comparable to experienced human readers, including during external validation. However, performance may decline in heterogeneous real-world populations for study-specific reasons not explained solely by scanner or protocol differences. No AI tool currently has regulatory clearance for autonomous use in axSpA, and applications remain largely confined to research and clinical trials.