Apply msmsulc to functionals

Joseph_Orr · January 16, 2024, 11:57pm

We’ve re-run our data through fmriprep 2023.2.0 in order to get the MSMsulc registrations, but this only generates *msmsulc_sphere.surf.gii files in anat. What is the best way to apply the transformation to functional data? I’ve used the HCPpipelines for fMRI surface registration, but its not very straightforward to determine all they are doing. Are the necessary files present to use the msm functions msmapplywarp or msmresample?

Steven · January 17, 2024, 12:01am

Hi @Joseph_Orr,

If you run with the --cifti-output 91k flag, MSM-based registration will be applied to your BOLD data o get it into fsLR 32k template space as used in HCP.

Best,
Steven

Joseph_Orr · January 17, 2024, 3:11am

Great thanks. I have multi-echo data that I subsequently run through tedana. Can you advise on whether or not there is an easy way to apply the registration to tedana corrected data?

effigies · January 17, 2024, 2:13pm

There is not currently an easy way. The CIFTI resampling portion of the workflow is here:

github.com

nipreps/fmriprep/blob/d85fb53007ec2b8f00624ec8d32658ca1bc3c78c/fmriprep/workflows/bold/base.py#L514-L617


      
          if config.workflow.cifti_output:
              from .resampling import (
                  init_bold_fsLR_resampling_wf,
                  init_bold_grayords_wf,
                  init_goodvoxels_bold_mask_wf,
              )
          
              bold_MNI6_wf = init_bold_volumetric_resample_wf(
                  metadata=all_metadata[0],
                  fieldmap_id=fieldmap_id if not multiecho else None,
                  omp_nthreads=omp_nthreads,
                  mem_gb=mem_gb,
                  jacobian='fmap-jacobian' not in config.workflow.ignore,
                  name='bold_MNI6_wf',
              )
          
              bold_fsLR_resampling_wf = init_bold_fsLR_resampling_wf(
                  grayord_density=config.workflow.cifti_output,
                  omp_nthreads=omp_nthreads,
                  mem_gb=mem_gb["resampled"],

This file has been truncated. show original

This relies on a BOLD series that is already sampled into alignment with the T1w image (future versions will resample directly from BOLD).

I will be working on tooling to do things like this in the nearish future. If you have any interest in this, there is a stand-alone tool that currently will collate transforms and resample to target volumes spaces: GitHub - nipreps/resampler: Single-shot resampling utility for nipreps derivatives

That does not handle multi-echo data (HMC and SDC should be ignored) or surfaces yet, but it is much slimmer than fMRIPrep as a place to work through things.