I’d really appreciate if someone might be able to offer some insight into the following questions.
I would like to use AFNI’s AlphaSim/3dClustSim on fMRI data processed with FSL. It is my understanding that the smoothness required as input to AlphaSim should refer to that estimated from the residual fMRI data (i.e. the “res4d.nii.gz” image). In order to estimate smoothness of my dataset, it seems I have two choices. I could use the FSL program, “smoothest,” or the AFNI program, “3dFWHMx.”
However, I am a bit confused as to why I get very different smoothness calculations from these two programs. 3dFWHMx consistently yields much larger FWHM kernel sizes than smoothest.
For example, when I run
> 3dFWHMx –input stats/res4d.nii.gz –mask mask.nii.gz
I get FWHMx values of x=10.342, y=10.6017, z=9.72374
On the other hand, when I run
> smoothest –r stats/res4d.nii.gz –d <degrees of freedom> -m mask.nii.gz
I get RESELS = 319.143, which corresponds to FWHM of 6.8337 (cubed root of 319)
Furthermore, when running "smoothest" on residual data, a flag for degrees of freedom is required. However, the different dof inputs produce the exact same “resels” value. If this value isn’t being used, could it reflect a possible bug?
Any insight would be much appreciated. Thanks!