FIX: Reverse transform order for multi-step

[mgxd]

Fixes #429

Alter the order when loading a composite transform that includes 2 composite transforms, like is generated if using --multi-step-reg. This is required when mapping from the target space to the source voxels.

nibabies/nibabies/interfaces/resampling.py

Lines 563 to 576 in 6599e40

	# Retrieve the RAS coordinates of the target space
	coordinates = nt.base.SpatialReference.factory(target).ndcoords.astype('f4').T
	# We will operate in voxel space, so get the source affine
	vox2ras = source.affine
	ras2vox = np.linalg.inv(vox2ras)
	# Transform RAS2RAS head motion transforms to VOX2VOX
	hmc_xfms = [ras2vox @ xfm.matrix @ vox2ras for xfm in hmc]
	# After removing the head-motion transforms, add a mapping from boldref
	# world space to voxels. This new transform maps from world coordinates
	# in the target space to voxel coordinates in the source space.
	ref2vox = nt.TransformChain(transform_list + [nt.Affine(ras2vox)])
	mapped_coordinates = ref2vox.map(coordinates)

It's not clear to me if this is something that needs to be accounted for in nitransforms, or if we should handle this unorthodox case internally here - would love hear any thoughts

[codecov]

Codecov Report

Attention: Patch coverage is 71.42857% with 2 lines in your changes missing coverage. Please review.

Project coverage is 56.12%. Comparing base (6599e40) to head (1e5083a).

Files with missing lines	Patch %	Lines
nibabies/utils/transforms.py	71.42%	2 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #430      +/-   ##
==========================================
+ Coverage   56.11%   56.12%   +0.01%     
==========================================
  Files          70       70              
  Lines        6690     6697       +7     
  Branches      832      834       +2     
==========================================
+ Hits         3754     3759       +5     
- Misses       2732     2734       +2     
  Partials      204      204              

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