The official implementation of the method proposed in Benou and Riklin-Raviv "DeepTract: A Probabilistic Deep Learning Framework for White Matter Fiber Tractography" https://arxiv.org/abs/1812.05129.
If you find this code useful in your research or publication, please cite the paper:
@inproceedings{benou2019deeptract,
title={Deeptract: A probabilistic deep learning framework for white matter fiber tractography},
author={Benou, Itay and Raviv, Tammy Riklin},
booktitle={International Conference on Medical Image Computing and Computer-Assisted Intervention},
pages={626--635},
year={2019},
organization={Springer}
}
- Clone/download repo.
- Edit the config.py file to configure the parameters according to the desired usage. Follow the comments above each parameter.
- Arrange the data: the DeepTract script expects a folder named "data" with the following folders/files structure:
data
--- dwi
--- <dwi_file>.nii
--- <bvecs_file>.bvecs
--- <bvals_file>.bvals
--- labels
--- <tractography_file>.trk
--- mask
--- <brain_mask_file>.nii
--- wm_mask
--- <white_matter_mask>.nii
- For training a new DeepTract model, run:
deeptract.py --train
For running streamline tractography using a trained DeepTract model, run:
deeptract.py --track