Interpolation of aparcaseg registered to T1w

[kaitj]

The registered segmentations to subject T1w (sub-{subject}_space-T1w_desc-aparcaseg_dseg.nii.gz) appear to be incorrectly interpolating the labels. Screenshot of example included at bottom of bug report. Unclear if this affects other transformed segmentations. Original segmentations are correctly labelled.

What version of fMRIPrep are you using?

21.0.0rc1; also replicated in 20.2.1
(issue seems to begin to appear around 20.x as far as I can tell, transformed segmentations were fine in 1.5.x)

What kind of installation are you using? Containers (Singularity, Docker), or "bare-metal"?

Singularity (invoked via Snakemake)

What is the exact command-line you used?

(pulled from Snakemake's Snakefile)

'singularity run -e {params.fmriprep_sif} '
' ${{SLURM_TMPDIR}}/bids_{wildcards.subject} ${{SLURM_TMPDIR}}/fmriprep_{wildcards.subject} participant '
' --participant_label {wildcards.subject} '
' --fs-license-file {params.fs_license} '
' -w ${{SLURM_TMPDIR}}/fmriprep_work_{wildcards.subject} '
' --n_cpus {threads} '
' --omp-nthreads {threads} '
' --output-spaces MNI152NLin6Asym MNI152NLin2009cAsym T1w '
' --cifti-output '
' --use-aroma '
' --output-layout bids '
' --notrack  && '
' cp -Rv ${{SLURM_TMPDIR}}/fmriprep_{wildcards.subject}/sub-{wildcards.subject} {output.fmriprep_out}'

Have you checked that your inputs are BIDS valid?

Yes

Did fMRIPrep generate the visual report for this particular subject? If yes, could you share it?

Yes (although the segmentations are not included, can still have include a visual report if needed)

Can you find some traces of the error reported in the visual report (at the bottom) or in crashfiles?

No

Are you reusing previously computed results (e.g., FreeSurfer, Anatomical derivatives, work directory of previous run)?

FreeSurfer 7.1 (ran separately). Also replicated with a clean run.

fMRIPrep log

If you have access to the output logged by fMRIPrep, please make sure to attach it as a text file to this issue.

List the steps you performed that revealed the bug to you.
Include any code samples. Enclose them in triple back-ticks (```)
Like this:

1. Opened subject T1w and sub-{subject}_space-T1w_desc-aparcaseg_dseg.nii.gz in FSLeyes
2. Changed color table from greyscale to FS cortical (also inspecting the labels for a few voxels)

[effigies]

Thanks for reporting this. Looks like I broke things in nipreps/smriprep#201. We'll need to release a fixed version of sMRIPrep for both the 0.7.x series and 0.8.x series and then we can release fixed versions of fMRIPrep.

[effigies]

20.2.5 is released.