CFNet

This repository contains the supported pytorch code and configuration files to reproduce of CFNet.

Parts of codes are borrowed from nn-UNet. For detailed configuration of the dataset, please refer to nn-UNet.

Environment

Please prepare an environment with Python 3.7, Pytorch 1.7.1, and Windows 10.

Dataset Preparation

Datasets can be acquired via following links:

• ACDC
• The Synapse multi-organ CT dataset
• Brain_tumor
• Heart

Dataset Set

After you have downloaded the datasets, you can follow the settings in nnUNet for path configurations and preprocessing procedures. Finally, your folders should be organized as follows:

./cfnet/
./DATASET/
  ├── CFNet_raw/
      ├── CFNet_raw_data/
          ├── Task01_ACDC/
              ├── imagesTr/
              ├── imagesTs/
              ├── labelsTr/
              ├── labelsTs/
              ├── dataset.json
          ├── Task02_Synapse/
              ├── imagesTr/
              ├── imagesTs/
              ├── labelsTr/
              ├── labelsTs/
              ├── dataset.json
          ├── Task03_tumor/
              ├── imagesTr/
              ├── imagesTs/
              ├── labelsTr/
              ├── labelsTs/
              ├── dataset.json
          ├── Task04_Heart/
              ├── imagesTr/
              ├── imagesTs/
              ├── labelsTr/
              ├── labelsTs/
              ├── dataset.json
      ├── CFNet_cropped_data/
  ├── CFNet_trained_models/
  ├── CFNet_preprocessed/

Preprocess Data

• CFNet_convert_decathlon_task -i D:\Codes\Medical_image\UploadGitHub\CFNet\DATASET\CFNet_raw\CFNet_raw_data
• CFNet_plan_and_preprocess -t 2

Functions of scripts

• Network architecture:
  • CFNet\cfnet\network_architecture\CFNet_acdc.py
  • CFNet\cfnet\network_architecture\CFNet_synapse.py
  • CFNet\cfnet\network_architecture\CFNet_tumor.py
  • CFNet\cfnet\network_architecture\CFNet_heart.py
• Trainer for dataset:
  • CFNet\cfnet\training\network_training\CFNetTrainerV2_cfnet_acdc.py
  • CFNet\cfnet\training\network_training\CFNetTrainerV2_cfnet_synapse.py
  • CFNet\cfnet\training\network_training\CFNetTrainerV2_cfnet_tumor.py
  • CFNet\cfnet\training\network_training\CFNetTrainerV2_cfnet_heart.py

Train Model

• python run_training.py 3d_fullres CFNetTrainerV2_cfnet_synapse 2 0

Test Model

• python predict.py -i D:\Codes\Medical_image\UploadGitHub\CFNet\DATASET\CFNet_raw\CFNet_raw_data\Task002_Synapse\imagesTs -o D:\Codes\Medical_image\UploadGitHub\CFNet\DATASET\CFNet_raw\CFNet_raw_data\Task002_Synapse\imagesTs_infer -m D:\Codes\Medical_image\UploadGitHub\CFNet\DATASET\CFNet_trained_models\CFNet\3d_fullres\Task002_Synapse\CFNetTrainerV2_CFNet_synapse__CFNetPlansv2.1 -f 0

• python CFNet/inference_synapse.py

Acknowledgements

This repository makes liberal use of code from:

• nnUNet
• nnFormer
• SSCFormer