Hi Mirtes,
You can indeed reduce the threshold, and depending on what kind of data you work on this will work. A downside is that you will increase your search volume in a rather unpredictable way - for example, non-brain areas may well have values over 0.5 of the SPM-calculated mean (muscle in fMRI; bone in PET with [18F] tracers; etc.etc.), and you could end up doing thousands of unneccessary t-tests in order to include those voxels you want.
I have found that a much neater way is to work with explicit masking, for example defining the whole brain volume in your stereotaxic space (usually MNI), to _ex_clude non-brain areas (obviously you have to check your normalisation carefully beforehand, and make the mask big enough to allow for normalisation uncertainty), and then making sure everything _within_ your mask gets _in_cluded by either specifying the lowest absolute threshold you're interested in, or a really low relative threshold (e.g. 0.1 or 0.2).
I'm afraid I don't know where you would specify this in spm_fmri_spm_ui.m...
Good luck,
Alexander
-----Original Message-----
From: SPM (Statistical Parametric Mapping) [mailto:[log in to unmask]]On
Behalf Of Mirtes Fortes
Sent: 16 February 2005 12:13
To: [log in to unmask]
Subject: [SPM] modify spm code mask threshold
Hi
I had some problems with the mask created in a fixed-effect multi-subject
analysis which excluded many important voxels. After looking for a solution in
old messages of this list I found a message wrote by Karl in that he suggests
to reduce the grey matter threshold before estimating the SPm.mat.
Is it sufficient to change directly the SPM code ( spm_fmri_spm_ui.m at line
413) which is originally:
%-Masking structure automatically set to 80% of mean
%=======================================================================
try
TH = g.*gSF*defaults.mask.thresh;
catch
TH = g.*gSF*0.8;
for example to
TH = g.*gSF*0.5;
or is it necessary to change other codes also?
Thanks in advance,
Mirtes
|