CIFTI run with Nibabies data collected MNIInfant for BOLD and MNI152NLin6Asym for T1w

[tsalo]

Summary

Per #759, XCP-D should collect the MNI152NLin6Asym NIFTI for the executive summary when the input-type is Nibabies and the --cifti flag is used. However, based on a bug report by @madisoth and @lundq163, it looks like the executive summary plotted a mean BOLD image in MNIInfant space and a T1w in MNI152NLin6Asym.

Additional details

• xcp_d version: 0.4.0
• Docker version:
• Singularity version:

What were you trying to do?

singularity run --cleanenv \
    xcp_d_unstable_07032023a.sif /data /out participant \
    --input-type nibabies \
    --cifti \
    --despike \
    --resource-monitor \
    --dcan-qc \
    -w /work \
    --omp-nthreads 3 \
    --warp-surfaces-native2std \
    -f 0.3 \
    -m \
    -vv \
    --participant-label ??????

What did you expect to happen?

What actually happened?

Reproducing the bug

[tsalo]

The fsLR surfaces will overlay on the MNI152NLin6Asym anatomical image, so we need an anatomical image in that space.

The executive summary also include plots in which the T1w edges are overlaid on top of the BOLD reference image, and vice versa, so they need to be in the same space. I think that there will be a transform to MNI152NLin6Asym as long as fMRIPrep is run with --cifti-output. Unfortunately, there's no guarantee that the BOLD data will be outputted in MNI152NLin6Asym, so either we require users to include it as an output space or we warp the BOLD data to that space...

The more I think about it, the more I think I need to warp the BOLD ref and NIFTI BOLD to MNI152NLin6Asym within XCP-D.

I can do this by adding an inverted parameter to get_std2bold_xfms, which will give the transforms from the BOLD image's space to MNI152NLin6Asym. Then I can apply that transform to the BOLD images I need in MNI space in the executive summary workflow.