Curie Brief
Turn on cookies to sign in
Signing in saves your progress to your Curie account. We can only do that with cookies on — turn them on to continue.
News
A deep learning system can now automatically detect inflammatory and structural lesions in sacroiliac joint MRIs with accuracy rivaling expert readers. Trained on clinical trial data from patients with axial spondyloarthritis, the AI performed especially well at identifying ankylosis. The findings could boost consistency and scalability in both clinical trials and routine care.
A fully automated deep learning pipeline developed by researchers at the University of Oxford can reliably detect both active and structural lesions in sacroiliac joint (SIJ) MRIs of patients with axial spondyloarthritis (axSpA) — and it performs on par with expert human readers. The two-stage system first delineates the left and right SIJs, then identifies five lesion types: bone marrow edema, erosions, fat lesions, sclerosis, and ankylosis.
The AI was trained on data from the MEASURE 1 clinical trial (132 patients) and validated on two independent phase 3 trial datasets — PREVENT (n=555) and SURPASS (n=414) — demonstrating strong generalizability across different patient populations and study settings.
By the Numbers:
Why it matters: Consistent, scalable MRI interpretation is a major bottleneck in both clinical trials and everyday rheumatology care. An AI system that matches expert accuracy could reduce variability, speed up diagnosis, and pave the way for broader use of imaging biomarkers in axSpA management.
