Hi Brandall:
The problem you are having relates, I think, to the fact that the SVC
is designed to provide a statistical assessment of the *intensity* of
voxel activation within your SV search space. It is not designed to
tell you whether a cluster within your search space is larger than
might be expected by chance (which would be a rather odd and perhaps
difficult thing to do anyway).
Thus, SPM delivers a table of significant foci within your search
space relating to intensity but still lists clusters at their actual
size - which can often leak beyond your search area.
Hope that makes sense (and is correct).
Your use of a SVC of 0.4 resels is a little concerning. It makes no
sense to use a SVC of smaller size than a complete resel - one resel
being the minimum search space that can be considered an independent
unit within your data set. There has been discussion here as to what
constitutes an acceptable size for a SVC. My 2c worth is that it
depends on the structure - I have recently used five resels in the
thalamus and lentiform and ten resels in cortical regions.
Best wishes,
Stuart.
---- Begin Original Message ----
From: "Brandall Y. Suyenobu" <[log in to unmask]>
Sent: Fri, 7 Dec 2001 13:00:27 -0800
To: [log in to unmask]
Subject: SVC: k vs. search volume difference
Hello SPMers-
I need some advice regarding a problem I have encountered with
interpretation of a SVC analysis: A SVC analysis of a first level
interaction yielded cluster-level results of corrected p=0.002, k=285,
uncorrected p=0.005 with a 12 x 12 x 12 search volume. In the
parameters
listed at the bottom of the SPM results page, the search volume was
listed
as: S=1728 mm^3 = 216 voxels = 0.4 resels. How is it that the
cluster
level results list a k value (285) that exceeds the search volume
voxels
(216)? I thought the k value was the number of voxels exceeding the
threshold within the specified search volume and therefore could not
exceed
the search volume in size. (I found that I used a search volume which
is
specifically recommended against
in http://www.mrc-cbu.cam.ac.uk/Imaging/vol_corr.html, but I did not
see
how this would result in the discrepancy between the cluster-level
results
k value and the search volume). Can someone explain?
With thanks,
Brandall Y. Suyenobu, Ph.D.
Staff Research Associate
Brain Imaging Group
UCLA/CURE Neuroenteric Disease Program
Veterans Administration Greater Los Angeles Healthcare System
11301 Wilshire Boulevard
Building 115, Room 223
Los Angeles, California 90073
Tel: (310) 478-3711 ext. 40580
Fax: (310) 794-2864
http://www.med.ucla.edu/Cure
http://ibs.med.ucla.edu
http://mindbody.med.ucla.edu
---- End Original Message ----
UPMC MR Research Center
PUH B-804
200 Lothrop Street
Pittsburgh
PA 15213
Sent by Medscape Mail: FREE Portable E-mail for Professionals on the Move
http://www.medscape.com
|