WSI Classification on TCGA lung cancer dataset (test)
96.53AccuracyMExD
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| MExDPre-trained Backbone=CTransPath (SRCL)2025.03 | 96.53 | 0.9813 | 96.51 | |
| MExDPre-trained Backbone=ViT (MoCo V3)2025.03 | 95.87 | 0.9827 | 95.87 | |
| DSMIL-LCFeature extraction=Patch-based, Scale=Multiple2020.11 | 95.71 | 0.9815 | — | |
| ACMILPre-trained Backbone=ViT (MoCo V3)2025.03 | 95.33 | 0.9774 | 95.09 | |
| DSMILFeature extraction=Patch-based, Scale=Single2020.11 | 94.76 | 0.9809 | — | |
| MambaMILPre-trained Backbone=ViT (MoCo V3)2025.03 | 94.76 | 0.9776 | 94.68 | |
| MambaMILPre-trained Backbone=CTransPath (SRCL)2025.03 | 94.48 | 0.9716 | 94.23 | |
| CLAM-SBPre-trained Backbone=ViT (MoCo V3)2025.03 | 94.38 | 0.9693 | 94.38 | |
| CLAM-SBPre-trained Backbone=CTransPath (SRCL)2025.03 | 94.37 | 0.969 | 94.17 | |
| IBMILPre-trained Backbone=ViT (MoCo V3)2025.03 | 94.29 | 0.9746 | 94.23 | |
| ACMILPre-trained Backbone=CTransPath (SRCL)2025.03 | 94.1 | 0.968 | 94.08 | |
| ABMILFeature extraction=Patch-based, Scale=Single2020.11 | 93.81 | 0.9765 | — | |
| MS-ABMILFeature extraction=Patch-based, Scale=Multiple2020.11 | 93.81 | 0.9792 | — | |
| CLAM-MBPre-trained Backbone=ViT (MoCo V3)2025.03 | 93.72 | 0.9735 | 93.7 | |
| IBMILPre-trained Backbone=CTransPath (SRCL)2025.03 | 93.62 | 0.9745 | 93.52 | |
| TransMILPre-trained Backbone=ViT (MoCo V3)2025.03 | 93.62 | 0.964 | 93.51 | |
| WiKGPre-trained Backbone=CTransPath (SRCL)2025.03 | 93.24 | 0.9595 | 93.23 | |
| CLAM-MBPre-trained Backbone=CTransPath (SRCL)2025.03 | 93.05 | 0.9713 | 93.04 | |
| DSMIL-LCFeature extraction=SimCLR, Scale=Multiple2020.11 | 92.86 | 0.9583 | — | |
| WiKGPre-trained Backbone=ViT (MoCo V3)2025.03 | 92.76 | 0.9752 | 92.73 | |
| MHIM-MILPre-trained Backbone=ViT (MoCo V3)2025.03 | 92.17 | 0.9495 | 90.48 | |
| DSMILFeature extraction=SimCLR, Scale=Single2020.11 | 91.9 | 0.9633 | — | |
| MHIM-MILPre-trained Backbone=CTransPath (SRCL)2025.03 | 91.68 | 0.9673 | 91.34 | |
| Mean-poolingFeature extraction=Patch-based, Scale=Single2020.11 | 90.96 | 0.9625 | — | |
| MS-MIL-RNNFeature extraction=Patch-based, Scale=Multiple2020.11 | 90.96 | 0.9561 | — | |
| TransMILPre-trained Backbone=CTransPath (SRCL)2025.03 | 90.95 | 0.9589 | 90.94 | |
| DSMILPre-trained Backbone=CTransPath (SRCL)2025.03 | 90.67 | 0.964 | 90.65 | |
| MIL-RNNFeature extraction=Patch-based, Scale=Single2020.11 | 90.48 | 0.9636 | — | |
| DSMILPre-trained Backbone=ViT (MoCo V3)2025.03 | 90.09 | 0.9583 | 90.23 | |
| ABMILFeature extraction=SimCLR, Scale=Single2020.11 | 90 | 0.9488 | — | |
| MS-ABMILFeature extraction=SimCLR, Scale=Multiple2020.11 | 90 | 0.9551 | — | |
| DTFD-MILPre-trained Backbone=ViT (MoCo V3)2025.03 | 89.72 | 0.946 | 89.67 | |
| ABMILPre-trained Backbone=CTransPath (SRCL)2025.03 | 89.71 | 0.9587 | 89.67 | |
| DTFD-MIL (MaxMinS)threshold=0.5, number of pseudo-bags=8, cross-validation protocol=4-fold2022.03 | 89.4 | 0.961 | 89.1 | |
| DTFD-MIL (AFS)threshold=0.5, number of pseudo-bags=8, cross-validation protocol=4-fold2022.03 | 89.1 | 0.951 | 88.3 | |
| DTFD-MIL (MAS)threshold=0.5, number of pseudo-bags=8, cross-validation protocol=4-fold2022.03 | 89.1 | 0.955 | 89 | |
| MS-MIL-RNNFeature extraction=SimCLR, Scale=Multiple2020.11 | 89.05 | 0.9213 | — | |
| DS-MILthreshold=0.5, cross-validation protocol=4-fold2022.03 | 88.8 | 0.939 | 87.6 | |
| Mean-poolingFeature extraction=SimCLR, Scale=Single2020.11 | 88.57 | 0.9369 | — | |
| Patch-based w/o MILFeature extraction=Patch-based, Scale=Single2020.11 | 88.57 | 0.9506 | — | |
| ABMILPre-trained Backbone=ViT (MoCo V3)2025.03 | 88.57 | 0.9249 | 88.41 | |
| DTFD-MILPre-trained Backbone=CTransPath (SRCL)2025.03 | 88.38 | 0.9498 | 88.31 | |
| TransMIL2021.06 | 88.35 | 0.9603 | — | |
| Trans-MILthreshold=0.5, cross-validation protocol=4-fold2022.03 | 88.3 | 0.949 | 87.6 | |
| FocAtt-MILBackbone=DenseNet, Feature Extraction=Transfer Learning (No fine-tuning)2021.06 | 88 | — | — | |
| CLAM-MBthreshold=0.5, cross-validation protocol=4-fold2022.03 | 87.8 | 0.949 | 87.4 | |
| CLAM-SBthreshold=0.5, cross-validation protocol=4-fold2022.03 | 87.5 | 0.944 | 86.4 | |
| Classic AB-MILthreshold=0.5, cross-validation protocol=4-fold2022.03 | 86.9 | 0.941 | 86.6 | |
| DTFD-MIL (MaxS)threshold=0.5, number of pseudo-bags=8, cross-validation protocol=4-fold2022.03 | 86.8 | 0.919 | 86.3 | |
| MIL-RNNFeature extraction=SimCLR, Scale=Single2020.11 | 86.19 | 0.9107 | — | |
| MIL-RNN2021.06 | 86.19 | 0.9107 | — | |
| Max-pooling2021.06 | 85.93 | 0.9463 | — | |
| Coudray et al.Backbone=DenseNet, Feature Extraction=Transfer Learning (No fine-tuning)2021.06 | 85 | — | — | |
| Kalra & Adnan et al.Backbone=DenseNet, Feature Extraction=Transfer Learning (No fine-tuning)2021.06 | 85 | — | — | |
| Max Poolingthreshold=0.5, cross-validation protocol=4-fold2022.03 | 84.6 | 0.901 | 83.3 | |
| RNN-MILthreshold=0.5, cross-validation protocol=4-fold2022.03 | 84.5 | 0.894 | 83.1 | |
| CLAM-MB2021.06 | 84.22 | 0.9377 | — | |
| Mean Poolingthreshold=0.5, cross-validation protocol=4-fold2022.03 | 83.3 | 0.901 | 80.9 | |
| Khosravi et al.Backbone=DenseNet, Feature Extraction=Transfer Learning (No fine-tuning)2021.06 | 83 | — | — | |
| Max-poolingFeature extraction=Patch-based, Scale=Single2020.11 | 82.86 | 0.8958 | — | |
| CLAM-SB2021.06 | 81.8 | 0.8818 | — | |
| Max-poolingFeature extraction=SimCLR, Scale=Single2020.11 | 80.88 | 0.9014 | — | |
| DSMIL2021.06 | 80.58 | 0.8925 | — | |
| ABMIL2021.06 | 77.19 | 0.8656 | — | |
| Yu et al.Backbone=DenseNet, Feature Extraction=Transfer Learning (No fine-tuning)2021.06 | 75 | — | — | |
| PT-MTA2021.06 | 73.79 | 0.8299 | — | |
| Mean-pooling2021.06 | 72.82 | 0.8401 | — |