ENH: Allow unwarp to apply only to references

[effigies]

This PR adds a ref_only mode to init_unwarp_wf that allows the same workflow to be used only on reference volumes.

This also changes the output reference volume to be derived from the input reference volume, rather than the average of the corrected time series.

[effigies]

This is intended to reduce the footprint of the EPI fit workflow. The corrected reference image will be used for T1w coregistration.

[codecov]

Codecov Report

Patch coverage: 85.71% and project coverage change: -0.03% ⚠️

Comparison is base (375dd7a) 81.99% compared to head (661ca8f) 81.97%.

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #385      +/-   ##
==========================================
- Coverage   81.99%   81.97%   -0.03%     
==========================================
  Files          26       26              
  Lines        2271     2274       +3     
  Branches      281      282       +1     
==========================================
+ Hits         1862     1864       +2     
  Misses        362      362              
- Partials       47       48       +1     

Flag	Coverage Δ
unittests	∅ <ø> (∅)

Flags with carried forward coverage won't be shown. Click here to find out more.

Files Changed	Coverage Δ
sdcflows/workflows/apply/correction.py	81.81% <85.71%> (-1.52%)	⬇️

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

[oesteban]

This changes the idea that a new reference is generated with the unwarped data, but that is probably is good overall, as the programmer will be required to generate the new reference explicitly with the undistorted data, if they don't want to directly use the unwarped reference.

Looks great to me :)

[oesteban]

I came back to check one thing - the distorted_ref should be a realigned file if it has more than 1 timepoints (I am currently working with sbrefs having 4 volumes)

[effigies]

Sorry, you mean distorted_ref should be averaged, in case it has multiple time points? I thought by the time we got to this, sbrefs had already been merged to a single volume.

[oesteban]

Sorry, you mean distorted_ref should be averaged, in case it has multiple time points? I thought by the time we got to this, sbrefs had already been merged to a single volume.

If that is the expectation, we should make it very clear that it has been averaged. However, resample would handle that just fine, allowing less data munging before correcting.

[effigies]

Closing in favor of #346. We will probably want to make averaging and brain extraction optional in a future PR, to avoid duplicating effort when only a timeseries is needed.