Functional connectivity post-processing options in native T1w and standard space

Software Support
1

Summary of what happened:

I finished running fMRIPrep (version 25.2.5) on around 10k participants from the UK Biobank dataset (output spaces: MNI152NLin2009cAsym res-1 and T1w native space). I want to use XCP-D for postprocessing/denoising, but it doesn’t support native space NIfTI output, which I need for functional connectivity analyses of small subcortical nuclei (hypothalamic subfields and pineal gland) using FreeSurfer segmentations.

My analysis requires:

  • Denoised BOLD in both MNI and native T1w space

  • Yeo 7 network parcellation (all Schaefer resolutions)

  • ALFF and ReHo maps

  • No smoothing, no global signal regression

Is there a standardized toolbox that supports both output spaces? If not, what is the recommended approach for applying fMRIPrep confounds in native space — HALFpipe, nilearn, or something else? I’m new to the field and want to use the most reproducible current approach.

Command used (and if a helper script was used, a link to the helper script or the command generated):

PASTE CODE HERE

Version:

fMRIPrep version 25.2.5

Environment (Docker, Singularity / Apptainer, custom installation):

apptainer

Data formatted according to a validatable standard? Please provide the output of the validator:

currently in BIDS format

PASTE VALIDATOR OUTPUT HERE

Relevant log outputs (up to 20 lines):

PASTE LOG OUTPUT HERE

Screenshots / relevant information:

2

Hi @ida-mehr and welcome to neurostars!

I’d personally recommend Nilearn if you need to stay in subject space. But you can also warp your nuclei ROIs to MNI and continue using XCP-D output.

Best,

Steven

3

Thanks for the welcome and quick reply Steven!

Since I need to be in native space + due to the small size of my ROIs, I’ll use nilearn then. One small follow-up question - given this, I have to choose which regressors to include manually correct? In general is there a field standard for this or is it more study specific?

My supervisor and I previously decided not to do global signal regression due to the nature of the analysis, however we haven’t made any other decisions regarding nuisance regressors.

Thanks for your help!!

4

Hi @ida-mehr,

No, there is no field standard. It largely depends on factors like the quality of your data and how many temporal degrees of freedom (i.e. volumes) you can model / can spare to lose with regression. You might also be interested in recent work about dangers of motion censoring: [2603.07380] Excessive data censoring in fMRI undermines individual precision and weakens brain-behavior associations

Best,

Steven