Incorporating computer vision on smart phone photographs into screening for inflammatory arthritis: results from an Indian patient cohort
Authors' conclusion
Computer vision distinguishes patients and controls using smartphone photographs, showing promise as a screening tool.
Study context
- Population
- Consecutive patients with early inflammatory arthritis and healthy controls, Indian cohort, examined by a rheumatologist800 hands analysed
- Comparison
- CNN on smartphone hand photographs for patient vs control and for joint inflammation (wrist, index finger proximal interphalangeal (IFPIP), middle finger proximal interphalangeal (MFPIP))
- Outcome
- Diagnostic performance of convolutional neural networks (CNNs)
Methodological detailsconsecutive patients, CNN (Inception-ResNet-v2) fine-tuned; 80% training, 20% test set
- Consecutive patients
- CNN (Inception-ResNet-v2) fine-tuned; 80% training, 20% test set
Performance
As reported in the abstract
| Test or model | AUC | Sensitivity | Specificity | Accuracy |
|---|---|---|---|---|
| Screening CNN | — not stated in the abstract | 98% | 99% | 99% |
| Wrist CNN | 0.86 | 92% | 72% | 75% |
| IFPIP CNN | 0.88 | 89% | 72% | 73% |
| MFPIP CNN | 0.87 | 91% | 70% | 71% |
AUC
Screening CNN— not stated in the abstract
Wrist CNN0.86
IFPIP CNN0.88
MFPIP CNN0.87
Sensitivity
Screening CNN98%
Wrist CNN92%
IFPIP CNN89%
MFPIP CNN91%
Specificity
Screening CNN99%
Wrist CNN72%
IFPIP CNN72%
MFPIP CNN70%
Accuracy
Screening CNN99%
Wrist CNN75%
IFPIP CNN73%
MFPIP CNN71%
— not stated in the abstract
Further findings
Inter-rheumatologist concordance
- Concordance between two rheumatologists
- 0.89
Joint involvement
- Wrist involved
- 173/400
- MFPIP involved
- 134
- IFPIP involved
- 128
Original publication
Incorporating computer vision on smart phone photographs into screening for inflammatory arthritis: results from an Indian patient cohort. Rheumatology (Oxford). Published 16 December 2024. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →