synthtorch

This package may not work with current versions of PyTorch and will not be supported.

This package contains deep neural network-based (pytorch) modules to synthesize magnetic resonance (MR) and computed tomography (CT) brain images. Synthesis is the procedure of learning the transformation that takes a specific contrast image to another estimate contrast.

For example, given a set of T1-weighted (T1-w) and T2-weighted (T2-w) images, we can learn the function that maps the intensities of the T1-w image to match that of the T2-w image via a UNet or other deep neural network architecture. In this package, we supply the framework and several models for this type of synthesis. See the Relevant Papers section (at the bottom of the README) for a non-exhaustive list of some papers relevant to the work in this package.

We also support a non-DNN-based synthesis package called synthit. There is also a seperate package to gather quality metrics of the synthesis result called synthqc.

** Note that this is an alpha release. If you have feedback or problems, please submit an issue (it is very appreciated) **

This package was developed by Jacob Reinhold and the other students and researchers of the Image Analysis and Communication Lab (IACL).

Link to main Gitlab Repository

Requirements

• matplotlib
• nibabel >= 2.3.1
• niftidataset >= 0.1.4
• numpy >= 1.15.4
• pillow >= 5.3.0
• torch >= 1.2.0
• torchvision >= 0.2.1

Installation

pip install git+git://github.com/jcreinhold/synthtorch.git

Tutorial

5 minute Overview

Jupyter Notebook example

In addition to the above small tutorial and example notebook, there is consolidated documentation here.

Singularity

You can build a singularity image from the docker image hosted on dockerhub or through singularity-hub via the following command:

singularity pull shub://jcreinhold/synthtorch:latest

Test Package

Unit tests can be run from the main directory as follows:

nosetests -v tests

Citation

If you use the synthtorch package in an academic paper, please use the following citation:

@misc{reinhold2019,
    author       = {Jacob Reinhold},
    title        = {{synthtorch}},
    year         = 2019,
    doi          = {10.5281/zenodo.2669612},
    version      = {0.3.2},
    publisher    = {Zenodo},
    url          = {https://doi.org/10.5281/zenodo.2669612}
}

Relevant Papers

[1] C. Zhao, A. Carass, J. Lee, Y. He, and J. L. Prince, “Whole Brain Segmentation and Labeling from CT Using Synthetic MR Images,” in MICCAI MLMI, vol. 10541, pp. 291–298, 2017.