Dear Jerry,
> The conventional wisdom for setting the FWHM for smoothing is 2-3 times the voxel size.
>
> How does one apply this criteria when voxels are anisotropic (e.g. 2mm x 2mm x4.5mm)?
If you want to apply this rule (e.g. 2 times the voxel size) to voxel
sizes of say 2mm x 2mm x 4.5 mm, you would choose a FWHM of [4 4 9] mm.
>
> In Matthew Brett's tutorial on smoothing, "An Introduction to Smoothing" (www.mrc-cbu.cam.ac.uk/Imaging/smoothing.html), he states that "... for a single subject experiment, where you want to detect (for example) a thalamic signal, which may be in the order of a few mm across, it would be wiser to use a very narrow smoothing, or even no smoothing."
>
> So, when the activation size is thought to be on the order of the voxel size, what is the best smoothing criteria to use:
> 2-3 times voxel size
> equal to voxel size
> no smoothing
>
If you think that the interesting part of your signal is really very
focal (subvoxel size), you should attempt to analyze your data without
smoothing. Another more explorative approach might be to analyze one
subject's data with various smoothing filters, look at the results,
choose the filter which you think is the 'best' one with respect to your
specific question and analyze other subjects' data using this filter
size (where the FWHM can also be 0 mm).
Hope this helps, Stefan
--
Stefan Kiebel
Functional Imaging Laboratory
Wellcome Dept. of Cognitive Neurology
12 Queen Square
WC1N 3BG London, UK
Tel.: +44-171-833-7478
FAX : -813-1420
email: [log in to unmask]
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|