Spondyloarthritis & psoriatic arthritis
Performances of machine learning algorithms in discriminating sacroiliitis features on MRI: a systematic review
Authors' conclusion
The performance of machine learning algorithms in discriminating sacroiliitis features on MRI varied owing to the high heterogeneity between studies and the small sample sizes, overfitting, and under-reporting issues of individual studies.
Study context
- Population
- Studies of machine learning discrimination of sacroiliitis features on MRI, published 2008 to 4 June 202310 studies
- Comparison
- Machine learning algorithms vs human physicians (radiologists) in discriminating sacroiliitis features on MRI
- Outcome
- Performance of ML algorithms in discriminating sacroiliitis features on MRI (AUC, sensitivity, specificity)
Methodological detailsforest plots, two independent reviewers, ASAS sacroiliitis criteria as ground truth in over half of studies; regions of interest manually extracted
- Forest plots
- Two independent reviewers
- ASAS sacroiliitis criteria as ground truth in over half of studies; regions of interest manually extracted
Pooled result
As reported in the abstract
Diagnostic performance of machine learning algorithms in discriminating sacroiliitis features on MRI
- AUC of machine learning algorithms (reported by all studies), range
- 0.76 to 0.99
- Sensitivity of machine learning algorithms, range
- 0.56 to 1.00
- Specificity of machine learning algorithms, range
- 0.67 to 1.00
- Radiologist sensitivity in the same cohort, range (comparable to machine learning)
- 0.67-1.00
- Radiologist specificity in the same cohort, range (comparable to machine learning)
- 0.78-1.00
Further pooled results
- More than half of studies at high risk of bias in the analysis domain, owing to small sample size or overfitting
Original publication
Performances of machine learning algorithms in discriminating sacroiliitis features on MRI: a systematic review. RMD Open. Published 23 November 2023. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →