Use ANTs to register BOLD to T1 with fmriprep

Hi Contributors,

Thanks for all the work you put into fmriprep, what a great tool! I’ve been preprocessing imaging data with it for a few months and have some questions.

For registration, particularly in younger populations, it seems that ANTs outperforms other registration methods and is quickly becoming the preferred method. I’m trying to use it to register from BOLD → T1.

If I understand correctly, fmriprep does not use ANTS for registration between BOLD → T1. By default, bbregister from freesurfer is used, and if bbregister is disabled flirt from fsl is used (Processing pipeline details — fmriprep version documentation).

However, fmriprep does use ANTs to calculate the registration between T1 → standard space (such as MNI), and to apply both transform matrices BOLD → T1, T1 → MNI in a single interpolation step.

Is there a way to use ANTs to register (BOLD → T1) with fmriprep?
Could this alternative transform matrix also be used when registering to standard space (BOLD → T1, T1 → MNI)?


Hi @loudcaterpillar and welcome to neurostars!

I do not think there is a way at the moment.

How young is the population you are studying? I think as long as they are old enough that T1w images have good contrast, then BBR should perform well. Have you tried running fMRIPrep and evaluating the registrations?

If you are interested in babies, then you may find the related and aptly-named Nibabies workflow to be more suitable for you.