Despike in MRIQC

mriqc
#1

Hi all,

I am currently using MRIQC to check imaging data quality. I find there is one option to include despike in MRIQC. Does it mean imaging data after MRIQC are “despiked”. I am suggested to not run despike before head motion estimation, which is included in fMRIPrep. So I wonder, does it matter that I put despike in MRIQC, then run fMRIPrep (it seems despike cannot be done in fMRIPrep) for pre-processing, and use confounds in generated tsv files for further data analysis? Any suggestion or help is welcome! Thanks in advance!

Best,

Shengjie

#2

MRIQC should not be used for preprocessing, only for quality assessment. As regards fMRIPrep, it is a longstanding feature request (https://github.com/poldracklab/fmriprep/issues/699). Contributions are welcome :slight_smile:

#3

Hi Esteban,

Many thanks for your being prompt in responding! I ran MRIQC with/without despike option on the same group of data yesterday and yes, MRIQC won’t modify original data. As for despike itself, sorry I am not an expert on this issue so I am afraid I cannot help much. Since now this feature is not included in fMRIPrep, if I really want to apply it, what is your suggestion? Should I run despike (currently I plan to use 3dDespike from AFNI) after fMRIPrep?

Best,
Shengjie

#4

Hi @Shengjie_Xu,

I think adding 3dDespike from AFNI as is will not be a good idea - you might be reintroducing artifacts later in regression as you would be establishing a “modular” pipeline (see https://www.biorxiv.org/content/10.1101/407676v1).

You could, though, generate one more regressor based on the RMS difference between signal before and after 3dDespike. @rastko should correct me if this is a bit crazy. If it is not, we should consider this for fMRIPrep.

Cheers,
Oscar