Summary and Schedule

This Data Carpentry lesson aims to introduce learners to the analysis of diffusion Magnetic Resonance Imaging (dMRI) data using primarily Python. Its target audience includes researchers willing to discover the dMRI field, as well as under-graduate or graduate students willing to broaden their skills from related fields, such as neuroscience, computational neuroscience, or medical image analysis.

Prerequisites for instructors

This lesson assumes that the instructors have solid knowledge about diffusion MRI, scientific programming tools used throughout the lesson (see the BIDS-dMRI Setup page), and teaching experience. If you are teaching this lesson in a workshop, please see the Instructor notes.

Prerequisites for learners

This lesson assumes that the learners have a basic understanding of how a signal or image is represented digitally, an entry-level knowledge about aspects linked to medical image analysis, and a minimal set of computing and programming skills to install the required tools and to follow the lesson (see the BIDS-dMRI Setup page). This lesson requires learners to have completed the Introduction to MRI and BIDS carpentry corresponding to the Neuroimaging Curriculum.

The actual schedule may vary slightly depending on the topics and exercises chosen by the instructor.

This lesson uses JupyterLab as a web-based interactive computational environment for learning. JupyterLab integrates a text editor, a terminal and Jupyter Notebooks through a single user-friendly interface, which will be used to run any code for the lesson. The JupyterLab instance and all of the necessary software have already been pre-installed on Binder. Users may choose to use this platform to get up and running much quicker, or may choose to run the notebooks locally. If running locally, there are several pieces of software that must be installed. Although the main components required are Python-based, there are a few additional non-Python tools required.

Instructors should allocate some time to have all components installed.

Internet

Binder requires users to have internet access to launch the Jupyter Notebooks. If users choose to run the notebooks locally, they will require internet access to download and install the software dependencies.

Binder

Binder enables the user to open the collection of notebooks in this lesson in a web-based executable and interactive environment. No additional software needs to be installed locally in this case; it suffices to click on the launch binder badge in the repository.

Local

If users choose to run the Jupyter notebooks locally, the following dependencies will need to be installed:

  • ANTs : used to register different anatomical data.
  • FSL : used for different data preprocessing steps.
  • DIPY : used for diffusion MRI data processing.
  • FURY : used for anatomical data visualisation purposes.
  • Matplotlib : used for data visualisation purposes.
  • Nilearn : used for anatomical data visualisation purposes.
  • osfclient : used to download the necessary data.
  • PyBIDS : used to check the data structure BIDS compliance.

Bash

The installation instructions require terminal application (bash, zsh, or others) skills.

Git

In order to run the notebooks locally, users will need to retrieve the code using Git or directly as a zipped file from the lesson’s GitHub repository.

Virtual environments

A virtual environment is a tool that helps to keep dependencies required by different projects separate by creating an isolated, self-contained directory tree containing a set of particular Python packages required for a particular version of Python, and a project. This allows one to have different Python versions or different versions of a given package within the same computer without running the risk of mixing incompatible versions of different packages, and hence eliminating the risk of affecting the development or execution of a project.

There are multiple package management tools (regardless of whether a virtual environment is used or not). These setup instructions will use pip.

For the setup purposes, it will be assumed that Python is already installed in the local environment. Note that only Python 3+ versions are supported.

Linux

ANTs and FSL are command line tools. Given the disparity and extent of the steps involved in their installation, users should follow the specific installation instructions in the corresponding official documentation pages.

The rest of the dependencies (DIPY, FURY, Matplotlib, Nilearn, osfclient, and PyBIDS) of the lesson are Python packages. These also rely on a number of other Python packages, which are automatically installed when installing the former.

In order to install the Python dependencies, it suffices to run:

BASH 

$ pip install -r requirements.txt

from the root of the repository folder.

Additionally, a helpful module to view anatomical slices have been included in the utils folder. To use this in python, run the following command:

BASH 

$ export PYTHONPATH=$PYTHONPATH:`realpath .`

Users may choose to run the notebooks using Jupyter or iPython. The Jupyter dependency can be installed by running:

BASH 

$ pip install jupyter

iPython can be installed through pip running:

BASH 

$ pip install ipython

Some Python distributions, such as the one provided by Anaconda, might ship these packages by default.

Test the installation

Test installation information for a package can be checked by running, for example:

BASH 

$ pip show dipy

Similarly, it can be checked that a given package can be imported in Python by running, for example:

BASH 

$ python
>>> import dipy

Alternatively, the package version can also be checked by running, for example:

BASH 

$ python
>>> import dipy
>>> print(dipy.__version__)

You can also see the packages and versions of all pip-installed dependencies by typing:

BASH 

$ pip freeze

In order to run the notebooks, the notebook server needs to be started. Once the current directory changed to the root of the code directory, the server is started running:

BASH 

$ ipython notebook

if using IPython, and running:

BASH 

jupyter-lab

if using JupyterLab.

In either case, the commands will print some information about the notebook server in the terminal, and a web browser will be opened to the URL of the web application (by default, http://127.0.0.1:8888). The users will be presented to the directory structure of the current directory, and they will be able to run the notebook of interest.

For additional information about Python setups besides the package manuals, users are encouraged to read the Programming with Python Carpentries lesson.

The data used in the lesson is hosted in OSF. It can be downloaded by running:

BASH 

$ osf -p cmq8a clone ./data

Notebooks expect them to be placed in the data folder that exists in the root of the repository.

Note that the above command clones the entire repository, which may be quite large and take a while to download. Alternatively, data from a single subject is available and can be downloaded by running:

BASH 

$ cd ./data
$ osf -p cmq8a fetch ds000221_subject/ds000221_sub-010006.zip 
$ unzip ds000221_sub-010006.zip