In March of 2000
(http://www.mailbase.ac.uk/lists/spm/2000-
03/0205.html),
Karl offered the following relative to smoothing.
"The smoothness should be 2-3 times the voxel size of
the data that
enters the estimation procedure (i.e. after spatial
normalization).
One can therefor subsample the data to smaller voxels
and then smooth with
a kernel that approximates the original voxel size. It is
important to
note that the smoothness is the post hoc smoothness of
the residual fields
(given in the SPM table footnotes). This smoothness
may be much greater
than the size of the smoothing kernel."
I recently analyzed fMRI data that has native voxel
dimensions of
3.75x3.75x5mm.
After realign & coregister, voxel dimensions were still
3.75x3.75x5mm.
Without applying any smoothing via spm_smooth.m,
SPM reports the
smoothness as (given in the SPM table footnotes) as:
FWHM 8.6x8.6x10mm = 2.3x2.3x2 voxels
Thus it seems that without applying any smoothing via
spm_smooth.m, my
data meet the smoothness criteria.
Is this to be expected?
My usual approach is to apply a smoothing kernel of
twice the native voxel
size (e.g.7.5x7.5x10) regardless of the smoothness
reported by SPM , but
if I understand Karls criteria, no smoothing need be
applied to this data
to accomplish the desired smoothness. Actually, if I apply a filter kernel
of 7.5x7.5x10mm to the data via spm_smooth.m, SPM
reports the smoothness
as: FWHM 48.9x52.8x60.2mm=13x14.1x12 voxels!!!!!!!
So, is it best to process the data initially without applying
any
smoothing, and then based upon the smoothness
reported by SPM decide
whether and how much smoothing to apply via
spm_smooth.m?
Jerry Allison, Ph.D.
Medical College of Georgia
L. Stephen Miller, Ph.D.
Associate Professor
Department of Psychology
The University of Georgia
Athens, GA 30602-3013 USA
TEL: (706) 542-1173
FAX: (706) 542-8048
email: [log in to unmask]
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|