Draft MEDIC dynamic distortion correction method (second attempt)

[tsalo]

Closes #36. An alternative to #435 that installs MEDIC as a dependency instead of implementing the whole tool as a series of interfaces and workflows.

Changes proposed:

• Install Julia in the Dockerfile.
• Add warpkit to dependencies.
• Add a MEDIC interface.
• Add a workflow to run MEDIC on a list of magnitude files and a list of phase files.

[tsalo]

I cannot figure out how to get the FieldmapWrangler to find MEDIC-style setups (i.e., complex-valued, multi-echo BOLD scans). 😕

[effigies]

@tsalo Around L508 in your current wrangler.py, I think you want to do something like:

        medic_entities = {**base_entities, **{'part': 'mag'}}
        has_magnitude = tuple()
        with suppress(ValueError):
            has_magnitude = layout.get(
                suffix='bold',
                **medic_entities,
            )
        for mag_img in has_magnitude:
            phase_img = layout.get(**{**mag_img.get_entities(), **{'part': 'phase'}})
            if not phase_img:
                continue
            phase_img = phase_img[
            try:
                e = fm.FieldmapEstimation(
                    [
                        fm.FieldmapFile(mag_img.path, metadata=mag_img.get_metadata()),
                        fm.FieldmapFile(phase_img.path, metadata=phase_img.get_metadata()),
                    ]
                )
            except (ValueError, TypeError) as err:
                _log_debug_estimator_fail(
                    logger, "potential MEDIC fieldmap", [mag_img, phase_img], layout.root, str(err)
                )
            else:
                _log_debug_estimation(logger, e, layout.root)
                estimators.append(e)

[codecov]

Codecov Report

Attention: Patch coverage is 37.63441% with 58 lines in your changes missing coverage. Please review.

Project coverage is 40.30%. Comparing base (656a51d) to head (b164fa3).

Files with missing lines	Patch %	Lines
sdcflows/workflows/fit/medic.py	28.00%	18 Missing ⚠️
sdcflows/utils/phasemanip.py	6.66%	14 Missing ⚠️
sdcflows/utils/wrangler.py	0.00%	14 Missing ⚠️
sdcflows/interfaces/fmap.py	74.28%	9 Missing ⚠️
sdcflows/fieldmaps.py	25.00%	3 Missing ⚠️

❗ There is a different number of reports uploaded between BASE (656a51d) and HEAD (b164fa3). Click for more details.

HEAD has 4 uploads less than BASE

Flag	BASE (656a51d)	HEAD (b164fa3)
	6	2

Additional details and impacted files

@@             Coverage Diff             @@
##           master     #438       +/-   ##
===========================================
- Coverage   83.75%   40.30%   -43.46%     
===========================================
  Files          32       33        +1     
  Lines        2831     2923       +92     
  Branches      381      387        +6     
===========================================
- Hits         2371     1178     -1193     
- Misses        390     1667     +1277     
- Partials       70       78        +8     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[tsalo]

Thanks @effigies!

[tsalo]

I created a MEDIC-compliant test dataset (dsD), but it seems like the tests in test_wrangler use skeletons. Should I drop the test dataset and just generate a skeleton in the test file?

[effigies]

This was probably expecting valid headers in the images.

[effigies]

Should I drop the test dataset and just generate a skeleton in the test file?

Yeah, seems like a good idea.