RF: Fix ITK warp conversion to nitransforms format #3300
Conversation
The reverse of the TransformFixedParameters should match the template dimensions. This fell through the cracks because the templates tested (MNI152 variants) all have the same X and Z dimensions. Only when testing on a template with distinct values for shape did this become evident. Additionally, this removes any `FIXED_PARAMS` checks, as that is template specific and more useful for debugging the initial implementation.
Co-authored-by: Chris Markiewicz <[email protected]>
Break the transformation steps down to be very explicit, with comments.
Codecov ReportAttention: Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## master #3300 +/- ##
==========================================
+ Coverage 73.71% 73.79% +0.08%
==========================================
Files 56 56
Lines 4269 4263 -6
==========================================
- Hits 3147 3146 -1
+ Misses 1122 1117 -5 ☔ View full report in Codecov by Sentry. |
There was a problem hiding this comment.
LGTM - I have not tested it, but can appreciate the improved clarity.
The only concern I have is this code is copied in a number of places (here, nibabies, resampler package) - we should focus some energy on having a single distribution and importing from there (maybe an ohbm hackathon project?)
|
Actually just looked at the MNI-resampled boldref in https://app.circleci.com/pipelines/github/nipreps/fmriprep/2204/workflows/b0c88b0f-7a65-4904-a5d4-7dc26b8233f4/jobs/11504/artifacts, and it's degraded relative to your PR and master. Let me try to understand this... |
|
Or, if you just want to add the |
|
Hmm. I loaded the file with my function and yours, and they seem identical. I'm going to re-run the test, in case it's just a one-off weirdness. |
|
Okay, it appears to have been a one-off bad boldmask, not a problem with this function at all. |
And yes. We should get this fixed up in nitransforms so we can stop having our own special loader. |
24.0.0 (June 17, 2024) New feature release in the 24.0.x series. This release is an incremental improvement on 23.2.x, with some fixes for bugs discovered in the updated workflow. New features include separation of HTML reports by session for subjects with many BOLD runs, a new ``--fs-no-resume`` option to improve interoperability with less typical FreeSurfer directories, such as those generated by longitudinal FreeSurfer or FastSurfer, and adoption of DatasetLinks and BIDS-URIs, to follow the recommendations of recent versions of BIDS. With thanks to Dimitri Papadopoulos, Basile Pinsard, Celine Provins, Taylor Salo and Wang Hao-Ting for their contributions! * FIX: Add "double" type to allowed DisplacementFieldTransform (#3287) * FIX: Require recent templateflow, select correct aparc dseg.tsv (#3256) * FIX: Ensure proper templates are retrieved with sloppy (#3251) * FIX: Delete summary from functional report when separated by sessions (#3223) * FIX: Support lists in bids filter file containing ``null`` or ``*`` (#3215) * FIX: Re-enable anat fasttrack for dataset without t1w (#3202) * ENH: Use BIDSURI in init_ds_boldmask_wf (#3297) * ENH: Add templateflow to DatasetLinks (#3267) * ENH: Track proximal sources of functional GIFTIs (#3263) * ENH: Support named derivative paths (#3264) * ENH: Track Sources for standard-space outputs (#3262) * ENH: Add --fs-no-resume option to reuse existing FreeSurfer outputs without resuming (#3142) * ENH: Use BIDS URIs to track Sources in sidecars (#3255) * ENH: Ignore unselected subjects in BIDSLayoutIndexer (#3236) * ENH: Add metadata for motion parameters (#3245) * ENH: Separate anatomical and functional reports per session for densely sampled dataset (#3191) * ENH: Leverage T2w if available for BOLD -> anat coregistration (#3208) * RF: Fix ITK warp conversion to nitransforms format (#3300) * RF: Load report assembler from nireports (#3177) * DOC: Clarify ``--dvars-spike-threshold`` uses standardized DVARS (#3205) * TST: Update test to reflect new report generation behavior (#3210) * STY: Manual conversions to f-strings (#3241) * STY: Apply ruff/pyupgrade rule UP031 (#3280) * STY: Lint and style check full repository (#3221) * STY: Adopt ruff for linting and formatting (#3206) * MNT: Pin libitk 5.3 and note dependencies (#3298) * MNT: Upgrade ruff pre-commit, add fixing checks (#3283) * MNT: Complete transition from flake8/black to ruff (#3279) * MNT: Apply Repo-Review suggestions (#3194) * MNT: Verbatim copy of Apache license 2.0 (#3259) * MNT: Bump cryptography from 41.0.7 to 42.0.4 (#3234) * MNT: Drop copyright year, unused dunder fields (#3247) * MNT: Update environment pins (#3226) * MNT: Bump codecov/codecov-action from 3 to 4 (#3219) * DOCKER: Restore mincinfo binary (#3249) * CI: Move to new circle machine tags (#3248) * CI: Avoid ruff warning (#3244) * CI: Pass ruff tests (#3243)
Alternative to (and builds on) #3296.
Closes #3296.