Medical Image Classification on MIMIC-CXR (test) (Fairness/Calibration)
0.153EOdds (Age)Worst-Group Equalized Odds Regularization
Evaluation results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Worst-Group Equalized Odds RegularizationBackbone=DenseNet-121, Validation-set tuning=true, Macro-averaging=true, Youden’s J statistic=true2026.05 | 0.153 | 0.174 | 0.054 | 0.127 | 0.897 | 0.885 | 0.966 | 0.916 | -0.007 | |
| JTTBackbone=DenseNet-121, Validation-set tuning=true, Macro-averaging=true, Youden’s J statistic=true2026.05 | 0.192 | 0.22 | 0.055 | 0.156 | 0.88 | 0.86 | 0.963 | 0.901 | -0.015 | |
| GDROBackbone=DenseNet-121, Validation-set tuning=true, Macro-averaging=true, Youden’s J statistic=true2026.05 | 0.201 | 0.226 | 0.062 | 0.163 | 0.884 | 0.852 | 0.954 | 0.896 | -0.008 | |
| FaMIBackbone=DenseNet-121, Validation-set tuning=true, Macro-averaging=true, Youden’s J statistic=true2026.05 | 0.208 | 0.222 | 0.066 | 0.165 | 0.866 | 0.857 | 0.956 | 0.893 | -0.022 | |
| AdvBackbone=DenseNet-121, Validation-set tuning=true, Macro-averaging=true, Youden’s J statistic=true2026.05 | 0.216 | 0.203 | 0.044 | 0.154 | 0.864 | 0.871 | 0.961 | 0.899 | -0.01 | |
| FairBatchBackbone=DenseNet-121, Validation-set tuning=true, Macro-averaging=true, Youden’s J statistic=true2026.05 | 0.239 | 0.254 | 0.093 | 0.195 | 0.824 | 0.821 | 0.956 | 0.867 | -0.009 | |
| BaselineBackbone=DenseNet-121, Validation-set tuning=true, Macro-averaging=true, Youden’s J statistic=true2026.05 | 0.335 | 0.316 | 0.112 | 0.254 | 0.792 | 0.817 | 0.937 | 0.849 | — |