List your company
ResearchFormerly WizWand

Medical Image Classification on RSNA (AUC)

94.5AUCMed-UniC (ViT-L/32)
57.47667.08876.786.312May 31, 2023Oct 28, 2023Mar 26, 2024Aug 24, 2024Jan 21, 2025Jun 20, 2025Nov 18, 2025
Updated 7mo ago

Evaluation results

MethodLinks
Med-UniC (ViT-L/32)Backbone=ViT-L/32, Training Data=100%, Protocol=Linear classification2023.0594.5
Med-UniC (ViT-L/32)Backbone=ViT-L/32, Training Data=10%, Protocol=Linear classification2023.0593.8
Med-UniC (ViT-B/16)Backbone=ViT-B/16, Training Data=100%, Protocol=Linear classification2023.0593.7
MRMBackbone=ViT, Training Data=100%, Protocol=Linear classification2023.0593.3
Med-UniC (ViT-B/16)Backbone=ViT-B/16, Training Data=10%, Protocol=Linear classification2023.0593.1
X-WINPretrain. data=372K CXRs + 32K CT volumes, Backbone=ViT-Large, Evaluation protocol=linear probing2025.1192.9
MRMBackbone=ViT, Training Data=10%, Protocol=Linear classification2023.0592.7
Med-UniC (ViT-L/32)Backbone=ViT-L/32, Training Data=1%, Protocol=Linear classification2023.0592.2
Med-UniC (ViT-B/16)Backbone=ViT-B/16, Training Data=1%, Protocol=Linear classification2023.0591.9
X-WINPretrain. data=372K CXRs + 32K CT volumes, Backbone=ViT-Base, Evaluation protocol=linear probing2025.1191.7
MRMBackbone=ViT, Training Data=1%, Protocol=Linear classification2023.0591.3
Med-UniC (ResNet-50)Backbone=ResNet-50, Training Data=100%, Protocol=Linear classification2023.0590.8
MGCA*Backbone=ViT-B/16, Training Data=100%, Protocol=Linear classification, disease-level annotations=true2023.0590.8
Med-UniC (ResNet-50)Backbone=ResNet-50, Training Data=10%, Protocol=Linear classification2023.0590.4
GLoRIA-MIMICBackbone=ResNet-50, Training Data=100%, Protocol=Linear classification2023.0590.2
MGCA*Backbone=ResNet-50, Training Data=100%, Protocol=Linear classification, disease-level annotations=true2023.0589.9
MGCA*Backbone=ViT-B/16, Training Data=10%, Protocol=Linear classification, disease-level annotations=true2023.0589.9
GLoRIA-MIMICBackbone=ResNet-50, Training Data=10%, Protocol=Linear classification2023.0589.4
MedKLIP*Backbone=ResNet-50, Training Data=100%, Protocol=Linear classification, disease-level annotations=true2023.0589.3
Ark+Pretrain. data=704K CXRs, Backbone=Swin-Base, Evaluation protocol=linear probing2025.1189.3
MGCA*Backbone=ResNet-50, Training Data=10%, Protocol=Linear classification, disease-level annotations=true2023.0589.1
Med-UniC (ResNet-50)Backbone=ResNet-50, Training Data=1%, Protocol=Linear classification2023.0589.1
MGCA*Backbone=ViT-B/16, Training Data=1%, Protocol=Linear classification, disease-level annotations=true2023.0589.1
GLoRIABackbone=ResNet-50, Training Data=100%, Protocol=Linear classification2023.0588.6
MGCA*Backbone=ResNet-50, Training Data=1%, Protocol=Linear classification, disease-level annotations=true2023.0588.6
GLoRIABackbone=ResNet-50, Training Data=10%, Protocol=Linear classification2023.0588
MedKLIP*Backbone=ResNet-50, Training Data=10%, Protocol=Linear classification, disease-level annotations=true2023.0588
MedKLIP*Backbone=ResNet-50, Training Data=1%, Protocol=Linear classification, disease-level annotations=true2023.0587.3
CheXFoundPretrain. data=987K CXRs, Backbone=ViT-Large, Evaluation protocol=linear probing2025.1187.2
GLoRIA-MIMICBackbone=ResNet-50, Training Data=1%, Protocol=Linear classification2023.0587
RAD-DINOPretrain. data=LVD-142M + 838K CXRs, Backbone=ViT-Base, Evaluation protocol=linear probing2025.1186.9
GLoRIABackbone=ResNet-50, Training Data=1%, Protocol=Linear classification2023.0586.1
CXR-AlignPretrain. data=325K CXRs w/ reports, Backbone=ViT-Base, Evaluation protocol=linear probing2025.1185.4
CARZeroPretrain. data=377K CXRs w/ reports, Backbone=ViT-Base, Evaluation protocol=linear probing2025.1183.3
CheXWorldPretrain. data=448K CXRs, Backbone=ViT-Base, Evaluation protocol=linear probing2025.1182.4
MaCoPretrain. data=377K CXRs w/ reports, Backbone=ViT-Base, Evaluation protocol=linear probing2025.1182.2
ConVIRTBackbone=ResNet-50, Training Data=100%, Protocol=Linear classification2023.0581.3
ConVIRTBackbone=ResNet-50, Training Data=10%, Protocol=Linear classification2023.0580.1
DINOv2Pretrain. data=LVD-142M, Backbone=ViT-Base, Evaluation protocol=linear probing2025.1179.8
CheXAgentPretrain. data=1.07M CXRs w/ text, Backbone=ViT-Large, Evaluation protocol=linear probing2025.1178.9
ConVIRTBackbone=ResNet-50, Training Data=1%, Protocol=Linear classification2023.0577.4
ImageNet InitTraining Data=100%, Protocol=Linear classification2023.0576.3
I-JEPAPretrain. data=ImageNet-1K, Backbone=ViT-Large, Evaluation protocol=linear probing2025.1175.4
ImageNet InitTraining Data=1%, Protocol=Linear classification2023.0574.9
ImageNet InitTraining Data=10%, Protocol=Linear classification2023.0574.5
Random InitTraining Data=100%, Protocol=Linear classification2023.0574.1
Random InitTraining Data=10%, Protocol=Linear classification2023.0569.4
Random InitTraining Data=1%, Protocol=Linear classification2023.0558.9