ENH: Add option to coregister BOLDs to "session" space#517
Conversation
workflow should only handle generation of a common, session boldref space. let downstream worry about boldref -> anat coregistration
TODO: handle func_fit_reports_wf Requires `boldref2anat_xfm` for some nodes, which may not be present at the time bold_fit_wf is run.
Codecov Report❌ Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## master #517 +/- ##
==========================================
+ Coverage 57.21% 57.53% +0.31%
==========================================
Files 72 73 +1
Lines 6977 7010 +33
Branches 860 860
==========================================
+ Hits 3992 4033 +41
+ Misses 2763 2756 -7
+ Partials 222 221 -1 ☔ View full report in Codecov by Sentry. 🚀 New features to boost your workflow:
|
mgxd
changed the title
| 'registrations (native -> MNIInfant -> template) and concatenate into a single xfm', | ||
| ) | ||
| g_baby.add_argument( | ||
| '--coreg-bolds', |
There was a problem hiding this comment.
Don't love this name. I'm kind of thinking something like --coreg-ref run/session, that allows us to potentially support other uses cases in the future.
Not sure that's a great alternative, but just to share where my brain's at and maybe that will trigger something for you. I'm also okay with keeping this and maybe/maybe not deprecating it in the future.
| fields=['fmap', 'fmap_ref', 'fmap_coeff', 'fmap_mask', 'sdc_method'], | ||
| key=fieldmap_id, | ||
| ), | ||
| name=f'fmap_select{i}', |
There was a problem hiding this comment.
For debugging purposes, could we use the same entity concatenation that we do in bold_<entities>_wf instead of i?
| ('coreg_boldref', 'inputnode.boldref'), | ||
| ('bold_mask', 'inputnode.bold_mask'), | ||
| ('motion_xfm', 'inputnode.motion_xfm'), | ||
| ('boldref2fmap_xfm', 'inputnode.boldref2fmap_xfm'), |
There was a problem hiding this comment.
Now that boldref is no longer a stand-in for "orig modulo motion", I wonder if we should rename boldref2fmap_xfm as orig2fmap_xfm globally. Right now, you could imagine that it's being registered to the intermediate boldref, which would require different handling.
| ('outputnode.boldref2anat_xfm', 'inputnode.boldref2anat_xfm'), | ||
| ('outputnode.motion_xfm', 'inputnode.motion_xfm'), | ||
| ('outputnode.boldref2fmap_xfm', 'inputnode.boldref2fmap_xfm'), | ||
| (inputnode, ds_bold_anat_wf, [ |
There was a problem hiding this comment.
Do you not need to pass orig2boldref_xfm here as well?
| ('bold_mask', 'inputnode.bold_mask'), | ||
| ('coreg_boldref', 'inputnode.hmc_boldref'), | ||
| ('motion_xfm', 'inputnode.motion_xfm'), | ||
| ('boldref2anat_xfm', 'inputnode.boldref2anat_xfm'), | ||
| ('dummy_scans', 'inputnode.skip_vols'), |
There was a problem hiding this comment.
It looks like we're missing the orig2boldref in the confounds workflow as well.
Though now I start to wonder if what we want are buffers where we populate the transform chains, like:
orig2ses_xfm = Merge(2) # HMC + orig2boldref
orig2anat_xfm = Merge(2) # orig2ses + boldref2anat
orig2std_xfm = Merge(2) # orig2anat + anat2stdThen workflows identify what specific chains they want, rather than the components that each has to sequence properly.
| 'boldref', | ||
| 'bold_mask', | ||
| 'motion_xfm', | ||
| 'orig2boldref_xfm', |
There was a problem hiding this comment.
native is supposed to be in orig space, not session. I think this workflow should be reverted, modulo the fieldmap changes.
| to_fmap_xfm = pe.Node( | ||
| niu.Merge(2), | ||
| name='to_fmap_xfm', | ||
| run_without_submitting=True, | ||
| mem_gb=config.DEFAULT_MEMORY_MIN_GB, | ||
| ) |
There was a problem hiding this comment.
Per out-of-band discussion, the fieldmap is registered directly to the run-space boldref, so orig2boldref_xfm should not be used here.
| else: | ||
| from niworkflows.data import load as nwf_load | ||
|
|
||
| boldref_buffer.inputs.orig2boldref_xfm = nwf_load('itkIdentityTransform.txt') |
There was a problem hiding this comment.
Is this even necessary? Generally leaving a transform empty has been a safe way of making generic workflows, e.g., omitting the boldref2anat_xfm resamples to boldref instead of anat.
2 participants
This PR adds in
--coreg-bolds, which generates a shared BOLD "session" reference space.If enabled, a template is created from each BOLD using
mri_robust_template.This change required a significant reorganization of the workflow, essentially stripping out much of the logic within
init_bold_wfinto the base workflow.One important caveat is this is not intended to be used if no SDC is to be applied and BOLD runs do not have a consistent phase encoding direction.