Confounds extraction from bold signal

,
1

Hello, I used fMRIPREP for preprocessing my bold images, the bold output is not perfectly registered to MNI space so I’m going to the registration again, just a quick question, Is that ok to remove the confounds from my signal and then apply for nonlinear registration on my bold images?

2

Hi @Mahsa,

Can you share the result of the original fMRIPrep registration, and your original fMRIPrep command? It is not ideal to perform registration again, as this involves data interpolation, which should be minimized.

Denoising, such as confound regression, should happen after all spatial normalization.

Best,
Steven

3

Here is my command, the result of registration for two different subject

Screen Shot 2023-07-04 at 10.29.34 AM
Screen Shot 2023-07-04 at 10.29.34 AM1920×778 69.3 KB

Screen Shot 2023-07-04 at 10.31.10 AM
Screen Shot 2023-07-04 at 10.31.10 AM1920×896 73 KB 

singularity run -B /scratch --cleanenv fmriprep-21.0.1.simg mahsa96/V1Bids /mahsa96/V1Bids/derivatives participant --participant-label 1 2\
        --skip-bids-validation \
        --fs-no-reconall\
        --md-only-boilerplate \
        --fs-license-file /mahsa96/V1Bids/derivatives/license.txt \
        --bold2t1w-dof 12\
        --output-spaces MNI152NLin6Asym\
        -w /scratch/mahsa96/mywork1\
        --stop-on-first-crash\
        --nthreads 16
4

Hi @Mahsa,

This is not recommended, see here. Please try rerunning by dropping this flag, enabling the FreeSurfer workflow. The boundary based registration workflow works better than volume based registration.

You can also try using the default DOF of 6.

Best,
Steven

5

I tried all the variations you mentioned before, but I don’t get a better registration result.