Thoracic disease classification on NIH ChestX-ray14 (test)
83.4AUROCViT-B-LR
Evaluation results
| Method | Links | |||||||||||||||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| ViT-B-LRArchitecture=ViT-B/16, Pre-training=X-rays (0.5M)2024.02 | 83.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MAEArchitecture=ViT-B/16, Pre-training=X-rays (0.5M)2024.02 | 83 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ViT-S-LRArchitecture=ViT-S/16, Pre-training=X-rays (0.3M)2024.02 | 82.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DN-121-LRArchitecture=DenseNet-121, Pre-training=ImageNet-1K2024.02 | 82.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MAEArchitecture=ViT-S/162024.02 | 82.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Kim et al.Architecture=DenseNet-1212024.02 | 82.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RN-50-LRArchitecture=ResNet-50, Pre-training=ImageNet-1K2024.02 | 82.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Hermoza et al.Architecture=DenseNet-1212024.02 | 82.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Liu et al.Architecture=DenseNet-1212024.02 | 81.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Ma et al.Architecture=DenseNet-121 (x2)2024.02 | 81.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Haghighi et al.Architecture=DenseNet-1212024.02 | 81.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Guan et al.Architecture=DenseNet-121, Pre-training=ImageNet-1K2024.02 | 81.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Seyyed et al.Architecture=DenseNet-1212024.02 | 81.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MAEArchitecture=DenseNet-121, Pre-training=X-rays (0.3M)2024.02 | 81.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Taslimi et al.Architecture=Swin Transformer2024.02 | 81 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Guendel et al.Architecture=DenseNet-1212024.02 | 80.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Baltruschat et al.Architecture=ResNet-502024.02 | 80.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MoCo v2Architecture=DenseNet-1212024.02 | 80.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| BaselineData type=real data2022.09 | 80.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PriCheXy-NetPrivacy parameter (mu)=0.0012022.09 | 80.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Tang et al.Architecture=ResNet-502024.02 | 80.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Ma et al.Architecture=ResNet-1012024.02 | 79.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PriCheXy-NetPrivacy parameter (mu)=0.0052022.09 | 79.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Wang et al.Architecture=ResNet-1522024.02 | 78.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Global-Local Contrastive Learning (augmented)Training Data=MIMIC-CXR + PadChest + ChestX-Ray14, Prompting Scheme=detailed2022.05 | 78.33 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Seibold et al.CLIP pretraining data=MIMIC, PadChest, ChestX-ray14, setting=zero-shot2023.03 | 78.33 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Label-SupervisedTraining Data=ChestX-Ray14, Supervision=Labels2022.05 | 76.47 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PriCheXy-NetPrivacy parameter (mu)=0.012022.09 | 76.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Yao et al.Architecture=ResNet & DenseNet2024.02 | 76.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Li et al.Architecture=ResNet-502024.02 | 75.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Wang et al.Architecture=ResNet-502024.02 | 74.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| XplainerCLIP pretraining data=MIMIC, setting=zero-shot2023.03 | 71.73 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Global-Local Contrastive LearningTraining Data=MIMIC-CXR, Prompting Scheme=detailed2022.05 | 71.23 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Seibold et al.CLIP pretraining data=MIMIC, setting=zero-shot2023.03 | 71.23 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| LoCoTraining Data=MIMIC-CXR, Prompting Scheme=detailed, Objective=Local2022.05 | 68.14 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SLIPTraining Data=MIMIC-CXR, Prompting Scheme=detailed2022.05 | 67.55 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| GloCoTraining Data=MIMIC-CXR, Prompting Scheme=detailed, Objective=Global2022.05 | 65.86 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CLIPTraining Data=MIMIC-CXR, Prompting Scheme=detailed2022.05 | 63.34 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MILNCElocalTraining Data=MIMIC-CXR, Prompting Scheme=detailed2022.05 | 63.06 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Privacy-Net2022.09 | 57.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DP-PixCell size (b)=82022.09 | 52.9 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DP-PixCell size (b)=42022.09 | 52.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DP-PixCell size (b)=22022.09 | 50.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DP-PixCell size (b)=12022.09 | 50 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Anatomy-XNetUtilizing Segmentation Masks=true, Input Resolution=224x2242021.06 | — | 0.9285 | 0.8442 | 0.9636 | 0.7171 | 0.7979 | 0.8604 | 0.8037 | 0.8306 | 0.9137 | 0.8091 | 0.899 | 0.8858 | 0.7709 | 0.8821 | 0.8505 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Anatomy-XNetUtilizing Segmentation Masks=true, Input Resolution=512x5122021.06 | — | 0.9433 | 0.8593 | 0.9457 | 0.7207 | 0.799 | 0.868 | 0.8378 | 0.8369 | 0.9138 | 0.8154 | 0.9025 | 0.8912 | 0.7748 | 0.9009 | 0.8578 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Arias-Garzón et al.Utilizing Segmentation Masks=true2021.06 | — | 0.8572 | 0.8168 | 0.8248 | 0.701 | 0.7767 | 0.8363 | 0.7892 | 0.8043 | 0.8893 | 0.8017 | 0.8771 | 0.8689 | 0.7507 | 0.8559 | 0.8179 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CheXNetUtilizing Segmentation Masks=false2021.06 | — | 0.9249 | 0.8219 | 0.9323 | 0.6894 | 0.7925 | 0.8307 | 0.7814 | 0.7795 | 0.8816 | 0.7542 | 0.8496 | 0.8268 | 0.7354 | 0.8513 | 0.818 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CLARiTy-S-16-224Image Size=224, NIH Training Set Size=70%2025.12 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 79.9 | |
| CLARiTy-S-16-512Image Size=512, NIH Training Set Size=70%2025.12 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 81.8 | |
| CRALUtilizing Segmentation Masks=false2021.06 | — | 0.908 | 0.83 | 0.917 | 0.702 | 0.778 | 0.834 | 0.773 | 0.781 | 0.88 | 0.754 | 0.875 | 0.829 | 0.729 | 0.857 | 0.8159 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DACNet2025.05 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 81.7 | 93.2 | 78.3 | 89.6 | 90.5 | 96.3 | 81.4 | 99.7 | 70.8 | 91.9 | 78.9 | 80.1 | 74 | 87.5 | — | |
| DualCheXNetUtilizing Segmentation Masks=false2021.06 | — | 0.942 | 0.837 | 0.912 | 0.705 | 0.796 | 0.838 | 0.796 | 0.784 | 0.888 | 0.746 | 0.852 | 0.831 | 0.727 | 0.876 | 0.823 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Ho et al.Utilizing Segmentation Masks=false2021.06 | — | 0.875 | 0.756 | 0.836 | 0.703 | 0.774 | 0.835 | 0.716 | 0.795 | 0.887 | 0.786 | 0.892 | 0.875 | 0.742 | 0.863 | 0.8097 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Keidar et al.Utilizing Segmentation Masks=true2021.06 | — | 0.9087 | 0.8147 | 0.918 | 0.706 | 0.7802 | 0.8393 | 0.7707 | 0.8064 | 0.9088 | 0.8043 | 0.892 | 0.8694 | 0.7653 | 0.8554 | 0.8314 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Li et al. (2022)Image Size=512, NIH Training Set Size=80%2025.12 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 82.9 | |
| LLAGNetUtilizing Segmentation Masks=false2021.06 | — | 0.939 | 0.832 | 0.916 | 0.703 | 0.798 | 0.841 | 0.79 | 0.783 | 0.885 | 0.754 | 0.851 | 0.834 | 0.729 | 0.878 | 0.8237 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Luo et al.Utilizing Segmentation Masks=false2021.06 | — | 0.9396 | 0.8381 | 0.9371 | 0.7184 | 0.8036 | 0.8376 | 0.7985 | 0.7891 | 0.9069 | 0.7681 | 0.861 | 0.8418 | 0.7419 | 0.9063 | 0.8349 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MANetUtilizing Segmentation Masks=true2021.06 | — | 0.8523 | 0.8282 | 0.921 | 0.7004 | 0.7682 | 0.8336 | 0.7776 | 0.8143 | 0.8935 | 0.8023 | 0.8856 | 0.863 | 0.7529 | 0.8546 | 0.8248 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Original CheXNetRelease Year=20172025.05 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 80.9 | 92.5 | 79 | 88.8 | 86.4 | 93.7 | 80.5 | 91.6 | 73.5 | 86.8 | 78 | 80.6 | 76.8 | 88.9 | — | |
| PCANImage Size=512, NIH Training Set Size=80%2025.12 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 83 | |
| Replicate CheXNet2025.05 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 76.2 | 92.2 | 74.6 | 86.4 | 88.3 | 85 | 76.6 | 92.5 | 67.3 | 82.4 | 64.6 | 75.6 | 65.6 | 82.7 | — | |
| RGTImage Size=224, NIH Training Set Size=70%2025.12 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 83.9 | |
| ThoraX-PriorNet-224Image Size=224, NIH Training Set Size=70%2025.12 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 84.4 | |
| ThoraX-PriorNet-512Image Size=512, NIH Training Set Size=70%2025.12 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 84.7 | |
| Transformer Model2025.05 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 77.4 | 89 | 78.9 | 87.6 | 85.7 | 82.8 | 77.2 | 87.2 | 70 | 78.3 | 67.3 | 76.6 | 71.3 | 82.1 | — | |
| Wang et al.Utilizing Segmentation Masks=false2021.06 | — | 0.933 | 0.838 | 0.938 | 0.71 | 0.791 | 0.834 | 0.777 | 0.779 | 0.895 | 0.759 | 0.855 | 0.836 | 0.737 | 0.877 | 0.826 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Wang et al. (2017)Image Size=1024, NIH Training Set Size=70%2025.12 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 74.5 | |
| Yan et al.Utilizing Segmentation Masks=false2021.06 | — | 0.9422 | 0.8326 | 0.9341 | 0.7095 | 0.8083 | 0.847 | 0.8105 | 0.7924 | 0.8814 | 0.7598 | 0.847 | 0.8415 | 0.7397 | 0.8759 | 0.8302 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — |