There is a default blurring option for afni_proc.py and I was curious what AFNI’s consensus is on blurring individual subject data. I use my errts files for input for ICA using FSL but all my pre-processing is done using afni_proc so if I don’t blur individual subjects, it goes into ICA ‘unsmoothed’.
It might be better to inquire with FSL how much blurring they
suggest before running the ICA, if any. I would expect not to
Ok - but as the afni_proc script is to create single subject processing scripts, and the default is to blur at 4.0mm, would your expertise say that individual subject data should be blurred? I get different results if I blur using the afni_proc script versus using 3dmerge -blur on the errts file output from afni_proc.
While the voxel size and registration method might affect the
blur size one chooses, nothing affects the choice more than
what one plans to do with the results.
In this case, your results are being handed to an FSL ICA
program, and it is likely that the software was designed with
some expectation about how much blur might be applied
before the software is run. It is quite possible that the
authors even expect no blurring to be applied.
Find out what the software expects, and let that influence
what you tell afni_proc.py to do.