Comparison of evaluation metrics of deep learning for imbalanced imaging data in osteoarthritis studies
Authors' conclusion
The commonly used ROC curve is not sufficiently informative, especially in the case of imbalanced data.
Study context
- Population
- Sagittal intermediate-weighted fat-suppressed knee MRIs with MOAKS readings from participants in the Osteoarthritis Initiative study
- Participants
- 2,467 participants2,996 knee MRIs
- Exposure
- Evaluation metrics (ROC and precision-recall curves) for deep learning detection of bone marrow lesions at sub-region, compartment and whole-knee levels
- Outcome
- Model performance for bone marrow lesions (BMLs) in a subregion with an extremely high imbalance ratio
Methodological detailsdeep learning, imbalanced imaging data
- Deep learning
- Imbalanced imaging data
Main finding
As reported in the abstract
ROC-AUC
0.84
PR-AUC
0.10
Sensitivity
0
Specificity
1
Observational data: associations, not effects.
Original publication
Comparison of evaluation metrics of deep learning for imbalanced imaging data in osteoarthritis studies. Osteoarthritis Cartilage. Published 19 May 2023. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →