Expanding from unilateral to bilateral: A robust deep learning-based approach for predicting radiographic osteoarthritis progression
Authors' conclusion
The DL model, constructed based on bilateral knee views, holds promise for enhancing the assessment of OA and demonstrates greater robustness during subsequent follow-up evaluations as compared with unilateral models.
Study context
- Population
- Participants in the Osteoarthritis Initiative with bilateral posteroanterior knee radiographs
- Participants
- 3,583 participants
- Comparison
- Bilateral-view deep learning (DL) model vs unilateral models
- Outcome
- Prediction of radiographic OA progression (AUC)
Methodological detailsdeep learning model development and validation, Osteoarthritis Initiative data, reader tests
- Deep learning model development and validation
- Osteoarthritis Initiative data
- Reader tests
Performance
As reported in the abstract
AUC for predicting OA progression
| Test or model | AUC, testing set 1 | AUC, testing set 2 | AUC |
|---|---|---|---|
| BikNet | 0.761 | 0.745 | — not stated in the abstract |
| ResNext | — not stated in the abstract | — not stated in the abstract | < 0.71 |
| DenseNet | — not stated in the abstract | — not stated in the abstract | < 0.71 |
AUC, testing set 1
BikNet0.761
ResNext— not stated in the abstract
DenseNet— not stated in the abstract
AUC, testing set 2
BikNet0.745
ResNext— not stated in the abstract
DenseNet— not stated in the abstract
AUC
BikNet— not stated in the abstract
ResNext< 0.71
DenseNet< 0.71
— not stated in the abstract
- BikNet vs ResNext and DenseNet: P < 0.001 — BikNet outperformed
Further findings
Clinicians' incident osteoarthritis (OA) prediction with vs without BikNet assistance (reader tests)
- Sensitivity, without to with BikNet
- from 28.1-63.2% to 42.1-68.4%
- Specificity, without to with BikNet
- from 57.4-83.4% to 64.1-87.5%
- Inter-observer reliability improved with BikNet
Original publication
Expanding from unilateral to bilateral: A robust deep learning-based approach for predicting radiographic osteoarthritis progression. Osteoarthritis Cartilage. Published 17 December 2023. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →