syn preprocessing `epi_reference_wf` does not distinguish between different matrix sizes about sdcflows HOT 6 CLOSED

Okay. I will rebase #312 on master, and we can aim to get it in for 2.5.0. Working on another bug at the moment, so might be tomorrow.

Not to shut down discussion, too quickly. Happy to consider other options, too.

from sdcflows.

#312 merged.

from sdcflows.

Just seen that this has been discussed elsewhere (#255, #312) I did check before I posted but clearly I didn't go far enough back...

Worth noting that no error was passed with respect to the reference workflow in my case, and it was just the downstream motion correction (using the reference image as the base) that flagged the problem.

from sdcflows.

Yes, we should deal with this. I see three options:

1. Split per-EPI (#312)
2. Group by shared parameters: (voxel sizes, TRT, echo time, PEDir)
3. Group by some shared parameters, use most common voxel size

Pushing me toward (1) is the fact that subjects might move within the scanner. If their head rotates around anything except the PE axis, then SyN will be working with out-of-plane blurs. If we're going to consider within-run head motion a problem to solve (#345), then we should not ignore it between-run.

In the future, we could consider methods to combine fieldmap estimates by shared parameters.

from sdcflows.

Although I was envisioning option 2 when I posted ( rotation between scans during anaesthetised scanning is pretty minimal and the biggest difference is usually a translation caused by field drift), I would also vote for option 1.

In its current draft, fprodents is estimating the distortion for each BOLD file separately when using PE-POLAR field maps, so I like the continuity across methods.

from sdcflows.

Tomorrow is fine; fprodents has a lot of other problems that need addressing simultaneously 😅

from sdcflows.