Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

LTA concatenation with NiTransforms producing wrong transforms #2145

Closed
oesteban opened this issue May 27, 2020 · 0 comments · Fixed by #2146
Closed

LTA concatenation with NiTransforms producing wrong transforms #2145

oesteban opened this issue May 27, 2020 · 0 comments · Fixed by #2146
Labels
bug CIFTI / Grayordinates critical effort: low Estimated low effort task freesurfer impact: high Estimated high impact task
Milestone
20.1.0

Comments

@oesteban
Copy link
Member

A problem using ConcatenateXFMs from the new niworkflows.interfaces.nitransforms module resulted in surface-sampled results looking like:

Screen Shot 2020-05-26 at 4 10 41 PM

because of the sampler receives the wrong transformation.

The problem is easily noticeable on the carpetplot:
image (1)

Thanks @mgxd for catching this and making all the necessary diagnosis.
@oesteban oesteban added the bug label May 27, 2020
@oesteban oesteban added this to the 20.1.0 milestone May 27, 2020
@oesteban oesteban added this to In progress in pipelines May 27, 2020
oesteban added a commit to oesteban/fmriprep that referenced this issue May 27, 2020
@oesteban
FIX: Generate proper LTA transform prior BOLD sampling on surfaces
f63af68 
This is a temporary patch before we go all the way in with NiTransforms
in the sampling of BOLD on surfaces.

The anatomical _fast-track_ required to expose the fsnative-to-T1w
transform in the derivatives folder (which we were already doing in ITK
format).

When fMRIPrep ran without the fast-track, then the LTA transform would
be directly passed in without conversions. The fast-track PR forced the
implementation to use the ITK version.

This, in conjunction with the little trick to stick the BOLD shape and
zooms into the LTA (i.e., using ``lta_concatenate`` with an identity
transform with those features, shape and zooms, as moving) resulted in
an overly complex workflow that I partially implemented with
NiTransforms.

This PR gets rid of the concatenation with identity trick, using
NiTransforms to generate a transform equivalent to the concatenated LTA
we used to generate before the fast-track was introduced.

Resolves: nipreps#2145
Assign: @mgxd
Milestone: 20.1.0
Related: nipreps#2118, nipreps#2041, nipreps#2121.
@oesteban oesteban mentioned this issue May 27, 2020
Merged
@oesteban oesteban added CIFTI / Grayordinates critical effort: low Estimated low effort task freesurfer impact: high Estimated high impact task labels May 27, 2020
pipelines automation moved this from In progress to Done May 27, 2020
@mgxd mgxd mentioned this issue May 28, 2020
Open
@oesteban oesteban removed this from Done in pipelines Sep 29, 2020
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees
No one assigned
Labels
bug CIFTI / Grayordinates critical effort: low Estimated low effort task freesurfer impact: high Estimated high impact task
Projects
None yet
Milestone
20.1.0
Development

Successfully merging a pull request may close this issue.

FIX: Generate proper LTA transform prior BOLD sampling on surfaces oesteban/fmriprep
1 participant
@oesteban