nnU-Net PanTS Model
nnU-Net model trained for pancreatic tumor segmentation, submitted to the PanTS leaderboard.
Usage
Download the model:
hf download edomerli/nnUNet-PanTS-regions --local-dir ./nnunet-pants-regions
Then use it as the model folder for nnUNetv2_predict_from_modelfolder.
See the code repo for full instructions.
Model details
- Framework: nnU-Net v2
- Configuration: 3d_fullres
- Trainer:
nnUNetTrainer_onlyMirror01_PanCrop - Plans:
nnUNetResEncUNetMPlansIso1x1x1 - Checkpoint:
checkpoint_final.pth - Code: nnUNet-PanTS-regions
Results (PanTS official in-distribution test set)
| Metric | Value |
|---|---|
| P-sensitivity @ 0.9 spec | 82.12% |
| P-specificity @ 0.9 spec | 90.13% |
| T-sensitivity | 78.15% |
| AUC | 0.912 |
| Pancreatic lesion DSC | 51.88% |
Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐ Ask for provider support