Running MRtrix3 using containers

From MRtrix version 3.0.3 onwards, official Docker and Singularity containers are provided for utilising MRtrix3 commands, enabling use of all MRtrix3 commands (including those that possess dependencies on other neuroimaging software packages) without necessitating any software installation on the user system.

FSL registration

The MRtrix3 containers include within them a subset of those commands that are part of FSL, as they are utilised within specific MRtrix3 Python scripts. Obtaining the FSL software typically necessitates registration. We therefore request that any users of MRtrix3 containers who have not previously registered as a user of the FSL software complete that process, as recognition of utilisation of their software.

Using Docker

Run terminal command

Non-GUI commands are typically executed as follows:

docker run --rm -it mrtrix3 <command>

(replacing “<command>” with the name of the command to be executed, along with any arguments / options to be provided to it)

If an MRtrix3 image has not been built on the local system, the most recent MRtrix3 Docker image will be automatically downloaded from [DockerHub](

Run GUI command

The following instructions have been shown to work on Linux:

xhost +local:root
docker run --rm -it --device /dev/dri/ -v /run:/run -v /tmp/.X11-unix:/tmp/.X11-unix -e DISPLAY=$DISPLAY -e XDG_RUNTIME_DIR=$XDG_RUNTIME_DIR -u $UID mrtrix3 mrview
xhost -local:root  # Run this when finished.

It may however be possible that you will need to modify these commands in order to operate without warning / error on your system.

Explicitly build image locally

As an alternative to downloading the image from DockerHub as described above, the following instruction can be run from a location in which the MRtrix3 source code has been cloned:

docker build --tag mrtrix3 .

Set DOCKER_BUILDKIT=1 to build parts of the Docker image in parallel, which can speed up build time. Use --build-arg MAKE_JOBS=4 to build MRtrix3 with 4 processors (can substitute this with any number of processors > 0); if omitted, MRtrix3 will be built using a single thread only.

Using Singularity

Build container natively

The following instruction can be run from the location in which the MRtrix3 source code has been cloned:

singularity build MRtrix3.sif Singularity

Build container from DockerHub

This command converts the Docker image as stored on DockerHub into a Singularity container stored on the user’s local system:

singularity build MRtrix3.sif docker://mrtrix3/mrtrix3:<version>

(Replace “<version>” with the specific version tag of MRtrix3 desired)

Run terminal command

Unlike Docker containers, where an explicit “docker run” command must be utilised to execute a command within a container, a Singularity image itself acts as an executable file that can be invoked directly:

MRtrix3.sif <command>

(replacing “<command>” with the name of the command to be executed, along with any arguments / options to be provided to it)

Run GUI command

The following basic usage has been shown to work on Linux:

singularity run -B /run MRtrix3.sif mrview

If you wish to utilise a clean environment when executing mrview, you will likely find that it is necessary to explicitly set the DISPLAY and XDG_RUNTIME_DIR environment variables. This could be done in a few different ways:

  1. Set environment variables that will be added to the clean environment of the container:

    singularity run --cleanenv -B /run MRtrix3.sif mrview
  1. Explicitly set those envvars during invocation (requires a relatively up-to-date Singularity):

    singularity run --cleanenv --env DISPLAY=$DISPLAY,XDG_RUNTIME_DIR=$XDG_RUNTIME_DIR -B /run MRtrix3.sif mrview
  1. Create a text file that specifies the environment variables to be set, and provide the path to that file at the command-line (requires a relatively up-to-date Singularity):

    singularity run --cleanenv --env-file ~/.singularity/mrview.conf -B /run MRtrix3.sif mrview

If you experience difficulties here with mrview, you may have better success if the Singularity container is built directly from the MRtrix3 source code using the definition file “Singularity” rather than converting from a Docker container or using a custom definition file.

If not using a clean environment, and you see the specific error:

Qt: Session management error: None of the authentication protocols specified are supported

This can be resolved by running: