DeLIS

This repository contains the source code to perform DeLIS a deep learning based intensity standardisation method.

The method is composed of three steps

Requirements

Some of the important requirements to perform the code are:

• python v3.6.4
• the repository of intensity normalization [1]
• Synthseg [2]

However, this requirements can be fulfil downloading the following docker image
docker pull emilyesme/tf_delis

Note: The DeLIS method use tensorflow backend to use the Synthseg. Therefore, it is necessary to have at less 10gb of ram in the host computer

Instructions

To use the DeLIS method it is necessary to have docker.io installed v19.03.8 or later and perform the following instructions

1. clone the repository
2. download the following image docker pull emilyesme/tf_delis
3. Move the run_intensity.py and run_delis.py to the data folder
4. To bind the docker container to the folder data, use
   docker run -it --mount type=bind,src=/path/to/repository/data,dst=/data emilyesme/tf_delis

Usage

To use the DeLIS method, here is an example with run_delis.py in the data folder:

python run_delis.py -b mri_image_1.nii.gz -o /data/

Instructions:

Usage:  python run_delis.py
        -b path to MRI T1-w scan (nii.gz file)
        -o output directory

References

[1] J. C. Reinhold, B. E. Dewey, A. Carass, and J. L. Prince, “Evaluating the impact of intensity normalization on MR image synthesis,” in Medical Imaging 2019: Image Processing, vol. 10949, p. 109493H, International Society for Optics and Photonics, 2019.
[2] B. Billot, D. Greve, K. Van Leemput, B. Fischl, J. E. Iglesias, and A. V. Dalca, “A learning strategy for contrast-agnostic mri segmentation,” arXiv preprint arXiv:2003.01995, 2020.