Motion parameter differences in FSL and FMRIPREP


I’ve been trying out two separate pipelines, one purely in FSL (preprocessing, stats), and one where I preprocess in fmriprep, and then perform my analyses in FSL. I’m noticing that the two motion parameters (both coming from FSL’s mcflirt), appear different in my design matrices.

Motion parameters fmriprep:

Motion parameters FSL:

I was curious to see why these appear different, and if this should be a concern for me?


Hi @dlevitas

Thank you for your question. There are a number of reasons why these could be different. A few could be the order of your processing steps and the parameters you used. Another could be the reference volume selection criteria. Differences in preprocessing can affect the realignment parameters. Are you using FEAT for the pure FSL preprocessing?

Correct, I use FEAT for the pure FSL preprocessing and leave the defaults (so middle volume is the reference during motion correction). I dug into the bold_hmc_wf and see that the mcflirt command uses -reffile, but I’m unsure which volume the reference file is exactly. Is it not the middle?

I believe this reference image is more than just the middle scan. This is the process for generating this ref image and motion correction estimation pipeline. This could help explain the differences seen in the mc parameters