dkmri 0.2.0

Reproducible and efficient diffusion kurtosis imaging in Python.

dkmri.py

dkmri.py stands for diffusion kurtosis magnetic resonance imaging in Python. It is a Python package for estimating diffusion and kurtosis tensors from diffusion-weighted magnetic resonance data. The estimation is performed using regularized non-linear optimization informed by a fully-connected feed-forward neural network that is trained to learn the mapping from data to kurtosis metrics. Details can be found in the arXiv preprint and source code.

This software can be used from the command line or in a Python interpreter.

• The command-line interface does not require any knowledge about Python.
• Python interface is for people comfortable with basic Python programming.

Installation

First, make sure you have installed Python.

If you just want to use the command-line interface, the recommended way of installing dkmri.py is to use pipx:

pipx install dkmri

pipx automatically creates an isolated environment in which the dependencies are installed.

If you want to use the Python interface, you can use pip (you should install dkmri.py in an isolated environment using venv or conda to avoid dependency issues):

pip install dkmri

Usage example

Command-line interface

The command for using dkmri.py is

dkmri.py data bvals bvecs optional-arguments

where data, bvals, and bvecs are the paths of the files containing the diffusion-weighted data, b-values, and b-vectors, and optional-arguments is where to define things such as which parameter maps to save.

For example, a command for computing a mean kurtosis map from data.nii.gz and saving it in mk.nii.gz could be

dkmri.py data.nii.gz bvals.txt bvecs.txt -mask mask.nii.gz -mk mk.nii.gz

To see a full description of the arguments, execute the following:

dkmri.py -h

Python interface

See the example notebook.

Support

If you have questions, found bugs, or need help, please open an issue on Github.

Citation

If you find this repository useful in work that leads to a scientific publication, please cite the arXiv preprint.