3D-ConvLSTMs-for-Monte-Carlo

This is the official repository for the article High-particle simulation of Monte-Carlo dose distribution with 3D ConvLSTMs presented in MICCAI 2021 (Strasbourg).

Dependencies

- numpy==1.17.4
- torchvision==0.9.1+cu111
- tqdm==4.50.0
- torch==1.8.1+cu111
- matplotlib==3.3.2
- asposestorage==1.0.2
- pytorch_ssim==0.1
- SimpleITK==2.1.0

Data

The Monte-Carlo dataset that we created and used for this research:

• is publically available here,
• has a comprehensive description here.

Training

To train a model, run python train.py in the folder where you wish to save training results.

You can change all the training parameters by adding specifications. For example python train.py --gpu 3 changes the identification number of your GPU to 3. All parameters can be found in the parse_args() function in utils.py.

You can find the weights of our best performing model (architecture: stacked 3D ConvLSTMs) from the article saved as best_val_model.pt.

Citation

If you find our project useful, please cite:

@inproceedings{martinot2021convlstm,
    title={High-particle simulation of Monte-Carlo dose distribution with 3D ConvLSTMs},
    author={Martinot, Sonia and Bus, Norbert and Vakalopoulou, Maria and Robert, Charlotte and Deutsch, Eric and Paragios, Nikos},
    booktitle={International Conference on Medical Image Computing and Computer-Assisted Intervention (MICCAI)},
    year={2021},
    organization={Springer}
}