FIX: Revise multi-echo reference generation, permitting using SBRefs too

[tsalo]

References #1745, nipreps/niworkflows#399, and nipreps/niworkflows#408.

Changes proposed in this pull request

• Currently, just tags a specific version of niworkflows to ensure backward compatibility, by request of @effigies.

Documentation that should be reviewed

None

[effigies]

Depends on: nipreps/niworkflows#408

[tsalo]

@effigies test_pytest failed due to what looks like a random connection error, unless I'm missing something and the failure is actually meaningful. Would you mind retriggering that test?

[effigies]

Done. Thanks for the heads up.

[effigies]

@tsalo I merged master, so you'll want to pull before doing anything else. Feel free to rebase, if you'd rather not have the pin and merge in the history,

[pvelasco]

Hi @tsalo,
Do you have a working Docker image of your multi-input-reference branch? I'm trying to test it on some ME-EPI data and I need a Singularity image.
I downloaded your branch, built a Docker image, tagged it as cbinyu/tsalo_fmriprep, uploaded it to DockerHub and finally did a singularity pull of that image.
When I ran it on some data, I got the following error:

Traceback (most recent call last):
  File "/usr/local/miniconda/bin/fmriprep", line 10, in <module>
    sys.exit(main())
  File "/usr/local/miniconda/lib/python3.7/site-packages/fmriprep/cli/run.py", line 312, in main
    opts = get_parser().parse_args()
  File "/usr/local/miniconda/lib/python3.7/site-packages/fmriprep/cli/run.py", line 50, in get_parser
    currentv = Version(__version__)
  File "/usr/local/miniconda/lib/python3.7/site-packages/packaging/version.py", line 277, in __init__
    raise InvalidVersion("Invalid version: '{0}'".format(version))
packaging.version.InvalidVersion: Invalid version: ''

So I was wondering if you had a working Docker image that I can singularity pull.
Thanks!

[tsalo]

@pvelasco I don't have a Docker image for it since I haven't touched this branch or the corresponding niworkflows branch in a while. However, I just resolved the current merge conflicts in nipreps/niworkflows#408 so if you switch the pinned niworkflows version in a current fMRIPrep version's (e.g., master's) setup.cfg to niworkflows @ git+https://github.com/tsalo/niworkflows.git@7a000cc7ebebffac50f1069ba0ecf8d005cced88 and build from that you should be good.

I'm reticent to do that here because I don't want to spam folks with a dummy commit.

[pvelasco]

Hi @tsalo,
I managed to get my a Docker image of your branch. I didn't notice that I had to use the VERSION argument in docker build.
Now the dataset is being processed. I'll report if the boldref image for multi-echo improves.

[tsalo]

@pvelasco Nothing should improve yet. I was waiting on nipreps/niworkflows#408 to be merged before trying to implement things in fMRIPrep, so the only change in this PR is the pinned version of nipreps in the setup file. That said, I can probably work on getting those changes implemented in the next few days.

[oesteban]

Okay, a side-by-side comparison of ROI plots seems reassuring:

• Unnoticeable changes for ds005
• Minimal changes for ds054 (none problematic). I attribute these to the fact that the SBRef is actually used now. Then, because subsequent runs of the workflow do not use the SBRef (Revise SBRef handling #1511), the final ROI plot does not really change.
• Finally, ds210 shows some changes but those are also minimal.

I think we can push this one in and keep working on #1511, #2182, #2174, and #2162.