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
- matplotlib
- nibabel
- niftidataset
- numpy
- torch
- torchvision
pip install git+git://github.com/jcreinhold/synthnn.git
In addition to the above small tutorial, there is consolidated documentation here.
You can build a singularity image from the docker image hosted on dockerhub via the following command:
singularity pull shub://jcreinhold/synthnn:latest
Unit tests can be run from the main directory as follows:
nosetests -v tests
If you use the synthnn package in an academic paper, please use the following citation:
@misc{reinhold2019,
author = {Jacob Reinhold},
title = {{synthnn}},
year = 2019,
doi = {10.5281/zenodo.2556299},
version = {0.1.4},
publisher = {Zenodo},
url = {https://doi.org/10.5281/zenodo.2556299}
}
[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.