Histopathology Classification on TCGA (test)
0.769F1 (Good prognosis)LDM synthetic augmentation
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| LDM synthetic augmentationAugmentation Strategy=LDM synthetic2026.01 | 0.769 | 0.675 | 0.526 | 0.835 | 0.68 | 0.701 | |
| Macenko stain augmentationAugmentation Strategy=Macenko stain normalization2026.01 | 0.76 | 0.654 | 0.389 | 0.801 | 0.622 | 0.651 | |
| Standard augmentationAugmentation Strategy=Standard2026.01 | 0.756 | 0.639 | 0.303 | 0.772 | 0.583 | 0.617 | |
| No augmentationAugmentation Strategy=None2026.01 | 0.754 | 0.655 | 0.359 | 0.797 | 0.611 | 0.641 | |
| Color jitteringAugmentation Strategy=Color jittering2026.01 | 0.753 | 0.632 | 0.293 | 0.767 | 0.576 | 0.611 | |
| LDM synthetic + Vahadane stain augmentationAugmentation Strategy=LDM synthetic + Vahadane2026.01 | 0.731 | 0.71 | 0.592 | 0.83 | 0.706 | 0.716 | |
| Geometric transformationAugmentation Strategy=Geometric transformation2026.01 | 0.73 | 0.639 | 0.411 | 0.791 | 0.617 | 0.643 | |
| Vahadane stain augmentationAugmentation Strategy=Vahadane stain normalization2026.01 | 0.726 | 0.672 | 0.486 | 0.775 | 0.647 | 0.665 | |
| LDM synthetic + Macenko stain augmentationAugmentation Strategy=LDM synthetic + Macenko2026.01 | 0.722 | 0.683 | 0.546 | 0.833 | 0.683 | 0.696 | |
| UNI + logistic regressionBackbone=UNI, Classifier=logistic regression2026.01 | 0.597 | 0.66 | 0.5 | 0.739 | 0.622 | 0.624 |