Cirrhosis Classification on D_histo
0.84AUCWeakly-supervised positional contrastive learning
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Weakly-supervised positional contrastive learningBackbone=TinyNet, Weak labels=true, Depth pos.=true2023.07 | 0.84 | — | |
| Weakly-supervised positional contrastive learningBackbone=ResNet-18, Weak labels=true, Depth pos.=true2023.07 | 0.84 | — | |
| depth-AwareBackbone=ResNet-18, Weak labels=false, Depth pos.=true2023.07 | 0.83 | — | |
| depth-AwareBackbone=TinyNet, Weak labels=false, Depth pos.=true2023.07 | 0.8 | — | |
| SupervisedBackbone=TinyNet, Weak labels=false, Depth pos.=false2023.07 | 0.79 | — | |
| SimCLRBackbone=ResNet-18, Weak labels=false, Depth pos.=false2023.07 | 0.79 | — | |
| BYOLBackbone=ResNet-18, Weak labels=false, Depth pos.=false2023.07 | 0.78 | — | |
| SupervisedBackbone=ResNet-18, Weak labels=false, Depth pos.=false2023.07 | 0.77 | — | |
| SupConBackbone=TinyNet, Weak labels=true, Depth pos.=false2023.07 | 0.76 | — | |
| SimCLRBackbone=TinyNet, Weak labels=false, Depth pos.=false2023.07 | 0.75 | — | |
| BYOLBackbone=TinyNet, Weak labels=false, Depth pos.=false2023.07 | 0.75 | — | |
| ImageNet (frozen representations)Backbone=ResNet-18, Weak labels=false, Depth pos.=false2023.07 | 0.72 | — | |
| None (random initialization)Backbone=ResNet-18, Weak labels=false, Depth pos.=false2023.07 | 0.69 | — | |
| SupConBackbone=ResNet-18, Weak labels=true, Depth pos.=false2023.07 | 0.69 | — | |
| None (random initialization)Backbone=TinyNet, Weak labels=false, Depth pos.=false2023.07 | 0.64 | — | |
| BYOLEncoder=Best of Tiny and ResNet2023.07 | — | 77 | |
| depth-AwareEncoder=Best of Tiny and ResNet2023.07 | — | 84 | |
| ImageNetEncoder=Best of Tiny and ResNet2023.07 | — | 74 | |
| None (random)Encoder=Best of Tiny and ResNet2023.07 | — | 71 | |
| Radiologists2023.07 | — | 82 | |
| SimCLREncoder=Best of Tiny and ResNet2023.07 | — | 78 | |
| SupConEncoder=Best of Tiny and ResNet2023.07 | — | 77 | |
| SupervisedEncoder=Best of Tiny and ResNet2023.07 | — | 78 | |
| WSP (Ours)Encoder=Best of Tiny and ResNet2023.07 | — | 85 |