This repository contains the code regarding our paper in MRM: Diffusion-derived intravoxel-incoherent motion anisotropy relates to CSF and blood flow
The aim of this repository is to enable you to use our IVIM-DTI-NET relatively easily on your own multi-b-value multi-directional diffusion-weighted data.
train_network.py --> trains a selfsupervised physics-informed neural network using multi-b-value multi-directional data, which you can provide yourself or can be downloaded from https://zenodo.org/records/12545278 (note that our code downloads this data automatically in the folder 'data').
After your network is trained, it is being saved in the folder 'trained_networks', and you can observe the corresponding loss curve in the folder 'plots'
Now, you can run predict_IVIM-DTI_parameters.py, which loads the trained network and predicts all IVIM-DTI model parameters. The IVIM-DTI parameter maps are saved in 'data/subject01/parammaps_IVIM-DTI-NET' as *.nii files having the same image space as the diffusion images. You can use a nifti viewer to see the paramater maps, (e.g., fsleyes).
To directly run the code, we added a '.yml' file which can be run in anaconda. To create a conda environment with the '.yml' file, enter the command in the terminal (e.g. Anaconda Powershell Prompt): conda env create -f environment.yml
This now creates an environment called 'ivimdti' that can be activated by: conda activate ivim
Paulien Voorter [email protected] | [email protected] | https://github.com/paulienvoorter
Note that this code is build upon previous repositories, and I would like to thank the authors for sharing their code:
June 2021 Oliver Gurney-Champion and Misha Kaandorp https://github.com/oliverchampion/IVIMNET
August 2019 Sebastiano Barbieri: https://github.com/sebbarb/deep_ivim