Hello,
I'm trying to do some cluster analysis, but I've been getting some abnormally high values for the spatial smoothness of my functional images using 3dFWHM. No spatial smoothing was applied. I'm using a 1-shot spiral sequence, matrix size 64x64, FOV 200 mm, 20 slices, slice thickness 3mm, thus the voxel dimension is 3.16mm x 3.16mm x 3.00mm. The output from 3dFWHM is as follows:
Program: 3dFWHM
Author: B. Douglas Ward
Initial Release: 20 February 1997
Latest Revision: 02 December 2002
count=81920
var =31643.213709
varxx=1825.244629 varyy=1978.071411 varzz=2515.451904
sx=9.242392 sy=8.872686 sz=7.448173
Gaussian filter widths:
sigmax = 9.24 FWHMx = 21.76
sigmay = 8.87 FWHMy = 20.89
sigmaz = 7.45 FWHMz = 17.54
These values force me to use a lower pthr and a large cluster size. Is it perhaps better to avoid an incorrect estimation of spatial smoothness and just lower the alpha? Or do these values seem reasonable?
I realize that this method, from Forman et al. 1995, is an improvement over the one given in Friston et al. 1991, but I don't know what all the caveats and assumptions are. I definitely plan on learning more about spatial autocorrelations over the next year, but I'm curious to know what others have discovered.