Overview

The source code for the paper "SAMmorph".

Setup

• Better using virtual environment to avoid conflicts. For example:

  conda create -n Segmet python=3.7
  conda activate Segmet
  pip install torch==1.9.1+cu111 torchvision==0.10.1+cu111 torchaudio==0.9.1 -f https://download.pytorch.org/whl/torch_stable.html
  pip install -r requirments.txt
  

  This demo is tested on Ubuntu 18.04 (Nvidia GPU required), but the training/testing code should be compatible with Windows as well.

Train

• Please check and set up the hyper-parameters in ./config/global_train_config.py and ./config/mpmrireg_train_config.py. Save your command line in a bash file, like those examples in ./scripts/mpmrireg.
• The source code can only use processed medical image data, the structure of the data files can be organized as follows:

|---./data/mureg_data/
        |---train
           |---mr_images
              |---case000000.nii.gz
              |---case000001.nii.gz
              .
              .
              |---case000064.nii.gz
           |---mr_labels
              |---...(structure same as above)
           |---us_images
              |---...(structure same as above)
           |---us_labels
              |---...(structure same as above)
        |---val
           |---...(structure same as above)
        |---test
              |---test_01.nii.gz
              |---test_02.nii.gz
              .
              .
              |---test_16.nii.gz
        |---SegData
           |---mr_labels
              |---case000007.nii.gz
              |---case000008.nii.gz
              .
              .
              |---case000072.nii.gz
              |---test_01.nii.gz
              |---test_02.nii.gz
              .
              .
              |---test_16.nii.gz
           |---us_labels
              |---...(structure same as above)
              
(If you use methods other than method Segmet, then the SegData folder is not necessary.)

• use following commandlines to repeat the experients in the paper

sh ./scripts/mpmrireg/[any of the bash file in it]

Test

Modify the method_math in the test.exe file to conduct testing on the trained model. Run the following:

python test.py

Reference Resources

Reference source for code section:

'voxelmorph', 'localnet', 'transmorph', 'MIDIR', 'CoTr', 'nnFormer', 
'PVTVNet', 'ViTVNet', 'mamba', 'CorrMLP', 'mpmrireg'

Feedbacks

• Please be free to create issues in this repo and to let me know if there're any problems. Thanks!