I am looking to run 3dClustSim for my current resting-state fMRI analysis. Before running 3dClustSim, I want to estimate the smoothness of the noise in my data using 3dFWHMx, however I am not sure which input to use.
From what I’ve seen in task-based experiments, the ‘errts’ files are typically used as the input as this represents the noise smoothness. However, in resting-state analyses the ‘errts’ file (or output of @Anaticor) is the signal of interest, not the noise. In my processing pipeline, I performed filtering, censoring, and blurring in 3dTproject after running @Anaticor, so my resulting connectivity maps are smoothed using a 6mm kernel. I then performed a seed to whole-brain correlation analysis for each subject (in subject space) and plan to run 3dMVM on the MNI warped z-score files.
Based on my processing pipeline, should I use my MNI warped z-score maps as my input to estimate smoothness in 3dFWHMx or should I be using a file prior to the correlation analysis (i.e. all_runs [subject space] prior to @Anaticor)?
Based on my processing pipeline, should I use my MNI warped z-score maps as my input to estimate
smoothness in 3dFWHMx or should I be using a file prior to the correlation analysis (i.e. all_runs
[subject space] prior to @Anaticor)?
Unfortunately there is no simple/easy answer for this. Currently there are a few assumptions embedded in the approaches to correcting for family-wise errors. And the average/uniform spatial correlation (estimated FWHM) is one of them. You may try both and see if they are close to each other.
When using 3dFWHMx to estimate the smoothness of
resting state data, it actually makes sense to just use
the pre-regression data, such as pb04.*.blur, say.
Note that the difference between epits and errts blur
estimates tends to be tiny anyway (less than 1%).
See this previous post.
The
National Institute of Mental Health (NIMH) is part of the National Institutes of
Health (NIH), a component of the U.S. Department of Health and Human
Services.