Provide a Docker image that includes pydeface

[SamGuay]

As mentioned on Neurostars, some people would like to use dcm2bids Docker container with pydeface.

To keep the image size to a minimum, I think we should take advantage of the solution mentioned in pydeface's #54.

I think if we go ahead and provide such container, it needs a distinctive tag and :latest should default to dcm2bids only.

[mateuszpawlik]

If it helps, here's the resources for the image I mentioned in https://github.com/poldracklab/pydeface/issues/54. I collected the flirt dependencies by trying to build it from sources.

Update: Using Conda is a better way now for installing FSL tools. See @SamGuay's comment below.

[SamGuay]

Thanks @mateuszpawlik for the info and the work towards the PR on pydeface to enable only flirt! I think we could follow a similar way of building a relatively small image using your approach in https://gitlab.com/ccns/neurocog/neurodataops/anc/containers/fsl-bet-flirt for which we can use the env below. Then we can shrink the image using conda-pack.

2 caveats still though:

1. pydeface 2.0.3 with PR#56 is only available on the devel branch for now.
2. Need to set FSLDIR="something random" because pydeface 2.0.3 initially checks for that env and triggers an error message if it is not set. However flirt runs without the $FSLDIR within a conda env..

name: dcm2bids-docker-env
channels:
  - conda-forge
  - https://fsl.fmrib.ox.ac.uk/fsldownloads/fslconda/public/
dependencies:
  - dcm2bids
  - dcm2niix
  - fsl-flirt
  - bids-validator
  - pip
  - pip:
    - git+https://github.com/poldracklab/pydeface.git@devel # since only v2.0.2 (on pypi)

[mateuszpawlik]

Glad you like it. I've changed to Conda because this is much simpler and officially supported way of installing the FSL tools. Feel free to open issues in my repo if you find anything interesting.

[arnaudbore]

This have been fixed and it's now part of the dev branch. We need to update the documentation and then it will be part of our next release.