Spondyloarthritis & psoriatic arthritis
Performance analysis of a deep-learning algorithm to detect the presence of inflammation in MRI of sacroiliac joints in patients with axial spondyloarthritis
Authors' conclusion
The algorithm enabled acceptable detection of inflammation according to the 2009 ASAS MRI definition in a large external validation cohort.
Study context
- Population
- Patients with nr-axSpA and r-axSpA from two prospective RCTs (RAPID-axSpA, n=152; C-OPTIMISE, n=579); mean age 34.2 years, 69.1% male; over 30 scanners, over 100 sites2 RCTs pooled
- Participants
- 731 patients
- Comparison
- Deep-learning algorithm vs central expert readings (2009 ASAS definition) for inflammation on sacroiliac joints (SIJ) MRI
- Outcome
- Detection of inflammation on SIJ MRI per 2009 ASAS definition vs central readings
- Follow-up
- Baseline
Methodological detailsexternal validation, blinded to clinical information and central expert readings
- External validation
- Blinded to clinical information and central expert readings
Performance
As reported in the abstract
Detection of inflammation on SIJ MRI (2009 ASAS definition), algorithm vs central expert readings
| Test or model | Sensitivity | Specificity | PPV | NPV | Cohen's kappa | Absolute agreement |
|---|---|---|---|---|---|---|
| Trained algorithm | 70% (95% CI 66% to 73%) | 81% (95% CI 78% to 84%) | 84% (95% CI 82% to 87%) | 64% (95% CI 61% to 68%) | 0.49 (95% CI 0.43 to 0.55) | 74% (95% CI 72% to 77%) |
Trained algorithm
Sensitivity70% (95% CI 66% to 73%)
Specificity81% (95% CI 78% to 84%)
PPV84% (95% CI 82% to 87%)
NPV64% (95% CI 61% to 68%)
Cohen's kappa0.49 (95% CI 0.43 to 0.55)
Absolute agreement74% (95% CI 72% to 77%)
Original publication
Performance analysis of a deep-learning algorithm to detect the presence of inflammation in MRI of sacroiliac joints in patients with axial spondyloarthritis. Ann Rheum Dis. Published 2 January 2025. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →