Dear Senichiro:
Hmm, are you getting your functional data at that resolution? In any case,
the nominal recommendation is to smooth at approximately twice your voxel
size. If you are truly getting high resolution functional data (voxel size
approx 1x1x1) then you could use a 2x2x2 smoothing kernel.
We get functional data at 64x64x32 slices for a voxel size of 3.75x3.75x4
mm. although smoothing with a 7 mm kernel is OK for individual results we
seem to get more robust results for groups if we smooth at 10x10x10. This
probably speaks to inaccuracies in normalization. it has also been noted on
the list that if we normalized and resliced our data with 3x3x3 mm voxels we
could probably get away with smoothing at 5x5x5 (i.e., we would satisfy the
gaussian random field assumptions although there are still the issues of
spatial normalization).
Finally Karl and others have noted before that the smoothing kernel that you
choose also depends on what you want to see in that according to the matched
filter theorem, you will enhance the signal of activations that are of the
order of your filter size. So large cortical activations might best be seen
with a larger size kernel, and smaller subcortical activations with a
smaller kernel. Of course it's also been pointed out that if one wants to go
"fishing" to find out the best filter sizes, that's best done on separate
data. Then one would take those results and apply it to the experimental
group. This is to avoid violating issues regarding repeated measures.
By the way search the list with keyword smoothing and I'm sure you'll find
more information.
Hope this helps,
Darren
-------------------------------------------------------------------
Darren R. Gitelman, M.D. E-mail: [log in to unmask]
Cognitive Neurology and Voice: (312) 908-9023
the Alzheimer's Disease Center Fax: (312) 908-8789
WWW: http://www.brain.nwu.edu
Northwestern Univ., 320 E. Superior St., Searle 11-470, Chicago, IL 60611
-------------------------------------------------------------------
> Dear SPM Experts:
>
> I am a SPM beginner and I have a simple question.
> In smoothing previous 256*192 matrix Analyze images (Signa 1.5T LX),
> what is the most suitable value of FWHM?
>
> Best regards,
>
> --
>
> Senichiro Kikuchi, M.D.
> EEG Lab.
> Dept. of Neuropsychiatry
> Gunma University School of Medicine
> 3-39-15 Shouwa-machi
> Maebashi-city
> Gunma pref.
> Japan
> Tel: +81-27-220-7111
> Fax: +81-27-220-8192
>
> [log in to unmask]
> [log in to unmask]
>
>
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|