JunMa11 / PETCTSeg

Automatic segmentation models for PET and CT scans

Automated Lesion Segmentation in Whole-body FDG-PET/CT: Solution to autoPET challenge

Introduction

The solution is based on the well-known nnU-Net. We make three modifications:

• using more data augmentations
• increasing the number of epochs to 1200
• DiceTopK loss function

The final model is the ensemble of 13 cross-validation models without testing-time augmentation.

Training

We train three groups cross-validation models

• Baseline model

nnUNet_train 3d_fullres nnUNetTrainerV2 taskid fold # fold in [0,1,2,3,4]

• more data agumentation

nnUNet_train 3d_fullres nnUNetTrainerV2_DA5 taskid fold # fold in [0,1,2,3,4]

• DiceTopK loss

nnUNet_train 3d_fullres nnUNetTrainerV2_DA5_DiceTopK10 taskid fold # fold in [0,1,2,3,4]

Inference

Donwload checkpoints: https://pan.baidu.com/s/1C3TaO0IVMXsBdSjAF-HMSg pw:4494 or https://drive.google.com/file/d/1gnSYN2Bn1sTDLXrTWOWGtJ3IUABkxhr_/view?usp=sharing

Run

docker build -t autopet_fighttumor .

Acknowledgements

• autoPET organizers: https://autopet.grand-challenge.org/
• nnUNet developers: https://github.com/MIC-DKFZ/nnUNet