ENH: Generate cortex mask in anat_fit_wf

[tsalo]

Based on @effigies' idea in nipreps/fmriprep#3491. This new output should make it possible to project data to the surface just using computed derivatives.

Changes proposed:

• Move roi output out of init_hcp_morphometrics_wf.
• Calculate that output in a new workflow, init_cortex_mask_wf.
• Add step to init_anat_fit_wf that calls init_cortex_mask_wf.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 75.21%. Comparing base (7679ad3) to head (04643a4).
⚠️ Report is 29 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #482      +/-   ##
==========================================
+ Coverage   70.82%   75.21%   +4.39%     
==========================================
  Files          25       25              
  Lines        2029     2062      +33     
  Branches      264      266       +2     
==========================================
+ Hits         1437     1551     +114     
+ Misses        537      444      -93     
- Partials       55       67      +12     

Flag	Coverage Δ
ds054	45.01% <ø> (?)

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

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

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[effigies]

If you want to use the thickness/midthickness GIFTIs as the sources, you're going to need to reorder things so that instead of going generate_surfaces -> surfaces_buffer -> datasink_workflows you do generate_surfaces -> datasink_workflows -> surfaces_buffer.

[effigies]

e.g.,

-            (gifti_surfaces_wf, surfaces_buffer, [
-                (f'outputnode.{surf}', surf) for surf in surfs
-            ]),
             (sourcefile_buffer, ds_surfaces_wf, [('source_files', 'inputnode.source_files')]),
             (gifti_surfaces_wf, ds_surfaces_wf, [
                 (f'outputnode.{surf}', f'inputnode.{surf}') for surf in surfs
             ]),
+            (ds_surfaces_wf, surfaces_buffer, [
+                (f'outputnode.{surf}', surf) for surf in surfs
+            ]),
         ])
         # fmt:on

and

-            (gifti_morph_wf, surfaces_buffer, [
-                (f'outputnode.{metric}', metric) for metric in metrics
-            ]),
             (sourcefile_buffer, ds_morph_wf, [('source_files', 'inputnode.source_files')]),
             (gifti_morph_wf, ds_morph_wf, [
                 (f'outputnode.{metric}', f'inputnode.{metric}') for metric in metrics
             ]),
+            (ds_morph_wf, surfaces_buffer, [
+                (f'outputnode.{metric}', metric) for metric in metrics
+            ]),

[tsalo]

I opened nipreps/niworkflows#962 to add the filename pattern for this mask. I don't know if desc-roi_mask.label.gii is the best name though. It seems a little too generic, but I don't know what would be better. desc-cortex also seems too generic.

[tsalo]

I thought sMRIPrep used niworkflows@master, and I waited until niworkflows had deployed. Am I missing something?

[tsalo]

Thanks for fixing that @effigies! I think this'll be good to go once CI passes, but I am still not sure desc-roi is the best name.

[mgxd]

personally i prefer desc-cortex over roi, since the sidecar metadata type is set to roi

[mgxd]

LGTM

[tsalo]

@effigies I think this is ready for one last review. Also, is there anyone else I should ping for a review?

[effigies]

@tsalo I wanted to suggest going back to iterables from explicit loops, but the changes would be too extensive to make a suggestion without testing. Please see the latest three commits and let me know if you're good with these changes.

[tsalo]

TBH I knew using iterables would be cleaner, but I never use them in my own code so I figured I'd mess it up. The only thing I want to check is that the Sources are correct in the output JSONs.

[tsalo]

@effigies thank you for refactoring the workflows. I looked at the desc-cortex_mask.json files in the ds005 derivatives and they looked good, so I think this is ready to merge.