ALFF/Resting state preprocessing - grand mean scaling

Hi Afni Experts,
I wanted to get your thoughts on the use of the use of grand mean scaling as it pertains to resting state fMRI and ALFF. From a paper from Calhoun and colleagues (http://inc.ucsd.edu/ica2001/075-calhoun.pdf) it seems that grand scaling is needed in resting state to account for gain which can change session to session. However, in previous posts it seems to hint that this step is needed in task-fMRI but not necessarily resting state fMRI since seed based analyses rely on shape profile as opposed to amplitude for correlation analysis. What are you current thoughts on this matter?

As it pertains to ALFF and mALFF, do you see this as a necessary step or could it introduce other unforeseen errors?

Thanks
Ajay

Hi, Ajay-

Well, scaling would be important, yes, because FMRI has no real unit-- but how scaling is actually done is equally important. I don’t see how grand mean scaling would be a good way to go for FMRI data.

As discussed in the “Units and Scaling” section of Gang’s paper here:
https://www.ncbi.nlm.nih.gov/pubmed/27729277
https://www.researchgate.net/publication/309099351_Is_the_Statistic_Value_All_We_Should_Care_about_in_Neuroimaging
… a better method would be to use voxelwise scaling, to result in a voxelwise BOLD %-signal change interpretation in your data. That would be more comparable both across subjects and across the brain.

Re. resting vs task for scaling-- the reason scaling might not often be used in resting is that time series are often not compared directly across subjects, as people mainly just calculate correlation coefficients within each subject’s brain; scaling doesn’t affect correlation coefficient calculation. However, for looking at ALFF changes (i.e., voxelwise magnitude changes), it would likely be quite important to scale, having a %-signal change that would be meaningful (again, both across subjects and across the brain).

–pt