ROI masking with featquery after fmriprep preprocessing and the Mumford workaround

Hi all,

I have preprocessed my data with fmriprep and did level-1 analyses in fsl. Then, after level-1 and before group analysis, I did the Mumford workaround to replace the mean_func files and all the .mat files. By that, if I understand correctly, I prevented further warping and processing of the images.

After doing the group analysis, I did ROI masking with fsl featquery. In order to do that, I regenerated the mean_func file using the command: updatefeatreg <path to the .feat folder of each subject> -gifs

After doing so I could do the ROI masking. BUT, now, the mean_func files in the stats folder and the reg_standard folder are not the same anymore. This is in contrast to Janet Mumford’s explicit warning that after the workaround, these files should be the same (I used the fslhd command to check it).

My question is- are the ROI analysis and whole-brain group analysis comparable in this case, even though different warp files are used? If not, how can I conduct both a whole-brain analysis in fsl using fmriprep preprocessed data AND do a fsl ROI analysis, using the same data?

Thank you very much,