Systemic sclerosis & Raynaud's
A deep learning system for quantitative assessment of microvascular abnormalities in nailfold capillary images
Authors' conclusion
Fully automated analysis using deep learning can achieve diagnostic performance at least as good as SSc experts, and is sufficiently robust to work with low-cost digital microscope images.
Study context
- Population
- Subjects with confirmed SSc (roughly half) and healthy controls or primary RP ('normal'); training set 111 subjects (group A)111 training subjects
- Comparison
- Automated deep learning nailfold capillaroscopy analysis vs SSc expert consensus
- Group Bn = 132
- Group Cn = 66
- Outcome
- Diagnostic performance (AUC, sensitivity/specificity) of automated SSc probability vs known clinical status (SSc vs 'normal')
Methodological detailsdeep learning, development and validation, fully automated image analysis
- Deep learning
- Development and validation
- Fully automated image analysis
Performance
As reported in the abstract
Discrimination of SSc vs 'normal'
| Test or model | AUC, median (90% CI) | AUC | Equal sensitivity/specificity | Sensitivity | Specificity |
|---|---|---|---|---|---|
| Group B | 97% (94-99%) | — not stated in the abstract | 91% (86-95%) | — not stated in the abstract | — not stated in the abstract |
| Group C | — not stated in the abstract | 95% (88-99%) | 89% (82-95%) | — not stated in the abstract | — not stated in the abstract |
| SSc expert consensus | — not stated in the abstract | — not stated in the abstract | — not stated in the abstract | 82% | 73% |
AUC, median (90% CI)
Group B97% (94-99%)
Group C— not stated in the abstract
SSc expert consensus— not stated in the abstract
AUC
Group B— not stated in the abstract
Group C95% (88-99%)
SSc expert consensus— not stated in the abstract
Equal sensitivity/specificity
Group B91% (86-95%)
Group C89% (82-95%)
SSc expert consensus— not stated in the abstract
Sensitivity
Group B— not stated in the abstract
Group C— not stated in the abstract
SSc expert consensus82%
Specificity
Group B— not stated in the abstract
Group C— not stated in the abstract
SSc expert consensus73%
— not stated in the abstract
Original publication
A deep learning system for quantitative assessment of microvascular abnormalities in nailfold capillary images. Rheumatology (Oxford). Published 18 January 2023. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →