On mBRSET, disentangling sensitive attributes from DR predictions improved DINOv2 AUROC by 2 points but dropped ConvNeXt V2 by 7 and Swin V2 by 3, with no consistent fairness gain.
Medical Image Analysis 95, 103209 (Jul 2024)
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Disentanglement and Assessment of Shortcuts in Ophthalmological Retinal Imaging Exams
On mBRSET, disentangling sensitive attributes from DR predictions improved DINOv2 AUROC by 2 points but dropped ConvNeXt V2 by 7 and Swin V2 by 3, with no consistent fairness gain.