Whole Slide Image classification on CAMELYON16 (test)
0.9981AUCHSA-NRL
Evaluation results
| Method | Links | |||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| HSA-NRLNoise ratio=10%2021.12 | 0.9981 | 0.9881 | — | — | — | — | 0.9882 | 0.9973 | 0.9773 | — | — | — | — | |
| AEMBackbone=ViT-L, Pre-training=SSL (UNI)2024.06 | 0.998 | 0.975 | — | — | — | — | — | — | — | — | — | — | — | |
| HSA-NRLNoise ratio=20%2021.12 | 0.9978 | 0.9869 | — | — | — | — | 0.9861 | 0.9961 | 0.974 | — | — | — | — | |
| ABMILBackbone=ViT-L, Pre-training=SSL (UNI)2024.06 | 0.996 | 0.968 | — | — | — | — | — | — | — | — | — | — | — | |
| HSA-NRLNoise ratio=30%2021.12 | 0.9957 | 0.983 | — | — | — | — | 0.9832 | 0.9984 | 0.968 | — | — | — | — | |
| HSA-NRLNoise ratio=40%2021.12 | 0.9951 | 0.9816 | — | — | — | — | 0.9817 | 0.9912 | 0.9722 | — | — | — | — | |
| OUSMNoise ratio=10%2021.12 | 0.9937 | 0.9806 | — | — | — | — | 0.9808 | 0.9955 | 0.9661 | — | — | — | — | |
| Co-tea.Noise ratio=10%2021.12 | 0.9934 | 0.9813 | — | — | — | — | 0.9816 | 0.9987 | 0.9645 | — | — | — | — | |
| INCVNoise ratio=20%2021.12 | 0.9933 | 0.9762 | — | — | — | — | 0.9765 | 0.9942 | 0.9587 | — | — | — | — | |
| Co-tea.Noise ratio=20%2021.12 | 0.9929 | 0.9753 | — | — | — | — | 0.9757 | 0.9929 | 0.9584 | — | — | — | — | |
| INCVNoise ratio=10%2021.12 | 0.9923 | 0.9784 | — | — | — | — | 0.9787 | 0.9957 | 0.9617 | — | — | — | — | |
| Full SupervisionProtocol=Full patch supervision (Upper bound)2023.03 | 0.992 | 0.967 | — | — | — | — | — | — | — | — | — | — | — | |
| SELFNoise ratio=10%2021.12 | 0.9902 | 0.9785 | — | — | — | — | 0.9789 | 0.9802 | 0.9753 | — | — | — | — | |
| Co-tea.Noise ratio=30%2021.12 | 0.9902 | 0.9744 | — | — | — | — | 0.9749 | 0.9968 | 0.953 | — | — | — | — | |
| JointNoise ratio=10%2021.12 | 0.9894 | 0.9753 | — | — | — | — | 0.9751 | 0.9689 | 0.9818 | — | — | — | — | |
| OUSMNoise ratio=20%2021.12 | 0.989 | 0.9652 | — | — | — | — | 0.9657 | 0.9833 | 0.9478 | — | — | — | — | |
| SELFNoise ratio=20%2021.12 | 0.9889 | 0.9749 | — | — | — | — | 0.9744 | 0.9788 | 0.9706 | — | — | — | — | |
| AEMBackbone=ViT-L, Pre-training=VLM (PathGen-CLIP)2024.06 | 0.988 | 0.967 | — | — | — | — | — | — | — | — | — | — | — | |
| JointNoise ratio=30%2021.12 | 0.9879 | 0.9603 | — | — | — | — | 0.961 | 0.9809 | 0.9406 | — | — | — | — | |
| JointNoise ratio=20%2021.12 | 0.9875 | 0.9649 | — | — | — | — | 0.9651 | 0.9713 | 0.9586 | — | — | — | — | |
| OUSMNoise ratio=30%2021.12 | 0.9873 | 0.9632 | — | — | — | — | 0.9643 | 0.9963 | 0.9322 | — | — | — | — | |
| SELFNoise ratio=30%2021.12 | 0.9872 | 0.97 | — | — | — | — | 0.9698 | 0.9756 | 0.9643 | — | — | — | — | |
| MHIM-MILBackbone=ViT-L, Pre-training=VLM (PathGen-CLIP)2024.06 | 0.984 | 0.946 | — | — | — | — | — | — | — | — | — | — | — | |
| ABMILBackbone=ViT-G, Pre-training=SSL (GigaPath)2024.06 | 0.984 | 0.978 | — | — | — | — | — | — | — | — | — | — | — | |
| FALFormerEncoder=CTransPath2024.07 | 0.983 | 0.958 | — | — | — | — | 0.9612 | 0.96 | 0.957 | — | — | — | — | |
| AEMBackbone=ViT-G, Pre-training=SSL (GigaPath)2024.06 | 0.982 | 0.981 | — | — | — | — | — | — | — | — | — | — | — | |
| LossAttnBackbone=ViT-L, Pre-training=VLM (PathGen-CLIP)2024.06 | 0.981 | 0.948 | — | — | — | — | — | — | — | — | — | — | — | |
| TransMILEncoder=CTransPath2024.07 | 0.978 | 0.933 | — | — | — | — | 0.938 | 0.942 | 0.926 | — | — | — | — | |
| DMNoise ratio=40%2021.12 | 0.9772 | 0.8839 | — | — | — | — | 0.8955 | 0.9984 | 0.793 | — | — | — | — | |
| Conveyance2026.05 | 0.977 | 0.905 | — | — | — | — | 0.933 | — | — | — | — | — | — | |
| INCVNoise ratio=30%2021.12 | 0.9764 | 0.9624 | — | — | — | — | 0.9628 | 0.9741 | 0.951 | — | — | — | — | |
| SELFNoise ratio=40%2021.12 | 0.9753 | 0.9589 | — | — | — | — | 0.9602 | 0.9678 | 0.9495 | — | — | — | — | |
| AEMBackbone=ViT-S, Pre-training=SSL (Lunit)2024.06 | 0.974 | 0.947 | — | — | — | — | — | — | — | — | — | — | — | |
| ABMILBackbone=ViT-L, Pre-training=VLM (PathGen-CLIP)2024.06 | 0.972 | 0.953 | — | — | — | — | — | — | — | — | — | — | — | |
| MHIM-MILBackbone=ViT-S, Pre-training=SSL (Lunit)2024.06 | 0.97 | 0.932 | — | — | — | — | — | — | — | — | — | — | — | |
| Clam-SBBackbone=ViT-S, Pre-training=SSL (Lunit)2024.06 | 0.969 | 0.925 | — | — | — | — | — | — | — | — | — | — | — | |
| DMNoise ratio=20%2021.12 | 0.969 | 0.9401 | — | — | — | — | 0.9398 | 0.9385 | 0.9417 | — | — | — | — | |
| Max-poolingBackbone=Fine-tuned ResNet-50 (Proposed)2023.03 | 0.969 | 0.927 | — | — | — | — | — | — | — | — | — | — | — | |
| TransMILBackbone=ViT-L, Pre-training=VLM (PathGen-CLIP)2024.06 | 0.968 | 0.951 | — | — | — | — | — | — | — | — | — | — | — | |
| CLAM-MBEncoder=CTransPath2024.07 | 0.968 | 0.935 | — | — | — | — | 0.938 | 0.931 | 0.938 | — | — | — | — | |
| DMNoise ratio=10%2021.12 | 0.967 | 0.9491 | — | — | — | — | 0.951 | 0.9915 | 0.9101 | — | — | — | — | |
| TransMILBackbone=Fine-tuned ResNet-50 (Proposed)2023.03 | 0.967 | 0.923 | — | — | — | — | — | — | — | — | — | — | — | |
| DSMILBackbone=ViT-S, Pre-training=SSL (Lunit)2024.06 | 0.966 | 0.943 | — | — | — | — | — | — | — | — | — | — | — | |
| CLAMInput type=image2025.03 | 0.966 | — | — | — | — | — | 0.915 | — | — | — | — | — | — | |
| OUSMNoise ratio=40%2021.12 | 0.9635 | 0.9132 | — | — | — | — | 0.9198 | 0.9864 | 0.8511 | — | — | — | — | |
| ILRABackbone=ViT-L, Pre-training=VLM (PathGen-CLIP)2024.06 | 0.963 | 0.929 | — | — | — | — | — | — | — | — | — | — | — | |
| INCVNoise ratio=40%2021.12 | 0.9626 | 0.9528 | — | — | — | — | 0.954 | 0.9806 | 0.9265 | — | — | — | — | |
| Co-tea.Noise ratio=40%2021.12 | 0.9625 | 0.9509 | — | — | — | — | 0.952 | 0.9756 | 0.9274 | — | — | — | — | |
| DTFD-MILBackbone=Fine-tuned ResNet-50 (Proposed)2023.03 | 0.962 | 0.921 | — | — | — | — | — | — | — | — | — | — | — | |
| AEMBackbone=ViT-B, Pre-training=VLM (CONCH)2024.06 | 0.961 | 0.942 | — | — | — | — | — | — | — | — | — | — | — | |
| Clam-SBBackbone=ViT-L, Pre-training=VLM (PathGen-CLIP)2024.06 | 0.96 | 0.941 | — | — | — | — | — | — | — | — | — | — | — | |
| CAMIL2023.05 | 0.959 | — | — | — | — | — | 0.917 | — | — | — | — | — | — | |
| CAMIL2026.05 | 0.959 | 0.881 | — | — | — | — | 0.917 | — | — | — | — | — | — | |
| NF-NetNoise ratio=30%2021.12 | 0.9581 | 0.8928 | — | — | — | — | 0.8961 | 0.925 | 0.8629 | — | — | — | — | |
| CLAM-SBBackbone=Fine-tuned ResNet-50 (Proposed)2023.03 | 0.956 | 0.911 | — | — | — | — | — | — | — | — | — | — | — | |
| PSMIL2026.05 | 0.956 | 0.921 | — | — | — | — | 0.922 | — | — | — | — | — | — | |
| JointNoise ratio=40%2021.12 | 0.9551 | 0.9156 | — | — | — | — | 0.9197 | 0.9694 | 0.8686 | — | — | — | — | |
| IBMILBackbone=ViT-S, Pre-training=SSL (Lunit)2024.06 | 0.954 | 0.912 | — | — | — | — | — | — | — | — | — | — | — | |
| IBMILBackbone=ViT-L, Pre-training=VLM (PathGen-CLIP)2024.06 | 0.953 | 0.935 | — | — | — | — | — | — | — | — | — | — | — | |
| CAMIL-L2023.05 | 0.953 | — | — | — | — | — | 0.91 | — | — | — | — | — | — | |
| DMNoise ratio=30%2021.12 | 0.9528 | 0.9308 | — | — | — | — | 0.9339 | 0.9774 | 0.8885 | — | — | — | — | |
| ABMILBackbone=ViT-B, Pre-training=VLM (CONCH)2024.06 | 0.952 | 0.932 | — | — | — | — | — | — | — | — | — | — | — | |
| TransMILInput type=image2025.03 | 0.95 | — | — | — | — | — | 0.938 | — | — | — | — | — | — | |
| TransMIL2023.05 | 0.95 | — | — | — | — | — | 0.905 | — | — | — | — | — | — | |
| CAMIL-G2023.05 | 0.95 | — | — | — | — | — | 0.891 | — | — | — | — | — | — | |
| DSMILBackbone=ViT-L, Pre-training=VLM (PathGen-CLIP)2024.06 | 0.949 | 0.895 | — | — | — | — | — | — | — | — | — | — | — | |
| DSMIL2023.05 | 0.949 | — | — | — | — | — | 0.874 | — | — | — | — | — | — | |
| DTFD-MIL (AFS)FLOPS=79.4M, Model Size=986.7K, number of pseudo-bags=52022.03 | 0.946 | 0.882 | — | — | — | — | 0.908 | — | — | — | — | — | — | |
| ABMILBackbone=ViT-S, Pre-training=SSL (Lunit)2024.06 | 0.945 | 0.914 | — | — | — | — | — | — | — | — | — | — | — | |
| DTFD-MIL (MAS)FLOPS=79.4M, Model Size=986.7K, number of pseudo-bags=52022.03 | 0.945 | 0.864 | — | — | — | — | 0.897 | — | — | — | — | — | — | |
| TransMILBackbone=ViT-S, Pre-training=SSL (Lunit)2024.06 | 0.943 | 0.922 | — | — | — | — | — | — | — | — | — | — | — | |
| DTFD-MIL (MaxMinS)FLOPS=80.1M, Model Size=986.7K, number of pseudo-bags=52022.03 | 0.941 | 0.865 | — | — | — | — | 0.899 | — | — | — | — | — | — | |
| DTFD-MIL2023.05 | 0.941 | — | — | — | — | — | 0.889 | — | — | — | — | — | — | |
| ILRABackbone=ViT-S, Pre-training=SSL (Lunit)2024.06 | 0.94 | 0.904 | — | — | — | — | — | — | — | — | — | — | — | |
| CLAM-MB2023.05 | 0.938 | — | — | — | — | — | 0.894 | — | — | — | — | — | — | |
| MR-RRTMILk=16, #P=2.042025.05 | 0.9371 | 0.8902 | — | — | — | — | 0.8982 | — | — | — | — | — | — | |
| NF-NetNoise ratio=10%2021.12 | 0.9353 | 0.897 | — | — | — | — | 0.9024 | 0.9516 | 0.8484 | — | — | — | — | |
| CLAM-SBEncoder=CTransPath2024.07 | 0.935 | 0.875 | — | — | — | — | 0.8837 | 0.881 | 0.87 | — | — | — | — | |
| FALFormerEncoder=ResNet502024.07 | 0.934 | 0.848 | — | — | — | — | 0.8605 | 0.86 | 0.84 | — | — | — | — | |
| MR-DGRMILk=8, #P=3.292025.05 | 0.934 | 0.8603 | — | — | — | — | 0.8673 | — | — | — | — | — | — | |
| CLAM-SB2023.05 | 0.933 | — | — | — | — | — | 0.877 | — | — | — | — | — | — | |
| FOCUSk=16, #P=86.92025.05 | 0.9322 | 0.8903 | — | — | — | — | 0.8983 | — | — | — | — | — | — | |
| DTFD-MILBackbone=Frozen IN-1K ResNet-502023.03 | 0.932 | 0.882 | — | — | — | — | — | — | — | — | — | — | — | |
| TransMILBackbone=ResNet50, Source=Reported in [23]2022.06 | 0.9309 | — | — | — | — | — | 0.8837 | — | — | — | — | — | — | |
| NF-NetNoise ratio=20%2021.12 | 0.9293 | 0.903 | — | — | — | — | 0.9072 | 0.9492 | 0.8612 | — | — | — | — | |
| LossAttnBackbone=ViT-S, Pre-training=SSL (Lunit)2024.06 | 0.928 | 0.908 | — | — | — | — | — | — | — | — | — | — | — | |
| MR-CATEk=16, #P=0.842025.05 | 0.9243 | 0.8673 | — | — | — | — | 0.8783 | — | — | — | — | — | — | |
| MR-DGRMILk=16, #P=3.292025.05 | 0.9235 | 0.8765 | — | — | — | — | 0.8875 | — | — | — | — | — | — | |
| CATEk=16, #P=1.932025.05 | 0.9223 | 0.8753 | — | — | — | — | 0.8852 | — | — | — | — | — | — | |
| GTPbatch size=22023.05 | 0.921 | — | — | — | — | — | 0.883 | — | — | — | — | — | — | |
| ProtoMILInput type=concept2025.03 | 0.918 | — | — | — | — | — | 0.907 | — | — | — | — | — | — | |
| NF-NetNoise ratio=40%2021.12 | 0.9172 | 0.8534 | — | — | — | — | 0.8553 | 0.8676 | 0.9396 | — | — | — | — | |
| KNN (Max)Backbone=Fine-tuned ResNet-50 (Proposed)2023.03 | 0.916 | 0.905 | — | — | — | — | — | — | — | — | — | — | — | |
| ProtoMILInput type=concept, Intervened=true2025.03 | 0.913 | — | — | — | — | — | 0.926 | — | — | — | — | — | — | |
| CLAM-SBEncoder=ResNet502024.07 | 0.91 | 0.849 | — | — | — | — | 0.8605 | 0.857 | 0.844 | — | — | — | — | |
| ABMILInput type=image2025.03 | 0.908 | — | — | — | — | — | 0.922 | — | — | — | — | — | — | |
| DTFD-MIL (MaxS)FLOPS=79.4M, Model Size=986.7K, number of pseudo-bags=52022.03 | 0.907 | 0.814 | — | — | — | — | 0.864 | — | — | — | — | — | — | |
| Trans-MILFLOPS=613.83M, Model Size=2.66M2022.03 | 0.906 | 0.797 | — | — | — | — | 0.858 | — | — | — | — | — | — | |
| DS-MILFLOPS=117.6M, Model Size=855.7K2022.03 | 0.899 | 0.815 | — | — | — | — | 0.856 | — | — | — | — | — | — | |
| FR-MILBackbone=ResNet18, Loss Formulation=Lbag2022.06 | 0.899 | — | — | — | — | — | 0.86 | — | — | — | — | — | — |