Unsuccessful Registration fmriprep 23.0.2

[zachladwig]

Hi, I previously had an issue with one of our data sets which I ended up chalking up to poor white/gray matter contrast in our multi-echo dataset and fmriprep not successfully using the sbref for registration.

Poor func/anatomical registration with fmriprep (21.0.2) on multi-echo data set

Hi! I am sometimes getting bad func/anatomical registration which results in low snr on parts of the brain where I wouldn’t expect to get low snr (like the lateral occipital lobe). Here is the fMRIprep command I used. singularity run --cleanenv -B /projects/:/projects/\ -B ${WORK_DIR}:/quest_scratch \ /projects/singularity_images/fmriprep-21.0.2.simg \ ${BIDS_DIR} \ ${BIDS_DIR}/${DERIVS_DIR} \ participant --participant-label ${subject} \ -w /quest_scratch --omp-nthread…

Here, I am using a different (single echo) dataset. And for some subjects I am getting terrible anatomical/functional registration. Here are a few examples. It seems to vary quite a bit by run and session. Here there is no sbref with better wm/gm contrast that I can use. I’m not sure what to do.

Any advice for what do do in this situation?

Functional boldref (in orange) laid on top of T1 (white)

Ses-1 Run 1

Ses-5 Run 5

[Steven]

Hi @zachladwig,

Would you mind giving the 23.2.0alpha a try? It can be pulled under the nipreps/fmriprep:next tag.

Best,
Steven

[jsein]

As suggested in the thread you reference:
you can try --bold2t1w-init header and/or --force-no-bbr to see if that improves your registration issue.

[zachladwig]

@Steven 23.2.0 had the same result

@jsein its strange. it actually used mri_coreg rather than bbreg but still the output was terrible (see above)

but, if I take the preprocessed intermediate file (after hmc) and then I do mri_coreg manually, it looks great. This should be the exact same thing right? I guess I am not sure what extra parameters they use on mri_coreg, i picked the most basic. I guess one solution is for me to manually register all of these from that intermediate step, but it seems like what you’re saying should have worked.

also @jsein I had played around with bold2t12-init header but I never felt like I understood exactly what it was doing. What exactly does that do?

[effigies]

@zachladwig

bold2t12-init header but I never felt like I understood exactly what it was doing. What exactly does that do?

Registration is a two-step: a bulk affine registration (mri_coreg) and a boundary-based registration refinement (bbregister). --bold2t1w-init header says to skip mri_coreg. --force-no-bbr says to skip bbregister. If you do both, fMRIPrep will not attempt to perform any coregistration.

[zachladwig]

@jsein

I re-ran it with force-no-bbr and the registration looks solid now for all of the runs/sessions! Even true for the runs which already didn’t use bbr (so used mri_coreg) in my last run and looked bad then. I don’t know why that is!

But maybe it will make sense for us to do force-no-bbr when we suspect our gray/white contrast is not good.