Jeff,
> Thanks for your help. I was going through your link below and the main
> part I am struggling with is how to get my or your estimate of
> intracerebral global intensity to be used by spm instead of their
> default (I am using spm2 but I assume the issues of integration in
> spm99 and spm2 are similar)
Scaling in SPM is subtle since it is applied to the file handles VY in
the design specification stage. (The file handles returned by spm_vol.m
contain information on the image and voxel dimensions, as well as
intensity scalefactors).
It seems like the easiest thing for you is to replace the call to
spm_global on line 494 of spm_fmri_spm_ui.m (in SPM99, or line 382 in
SPM2) with your own global function.
Another option is to pre-scale your images (somehow) and then use no
scaling. The following rudimentary script for SPM99 removes scaling:
%======================================================================
% Hack to turn off grand mean scaling (and maybe masking)
clear
load SPMcfg
% Turn off grand mean scaling
xGX.sGMsca = '<none>';
xGX.GM = [];
xGX.gSF = ones(size(xGX.gSF));
% % Note: To turn off masking, uncomment these lines
% Turn off masking for positive data
% xM.T = zeros(size(xM.T));
% xM.TH = zeros(size(xM.TH));
% Refresh VY to eliminate any scaling
VY = spm_vol(P);
save SPMcfg
%======================================================================
After defining the analysis but before estimating, run this script in
the anaysis directory. It will turn off scaling, and, if you
uncomment the associated lines, masking. The key thing here is that
you have to reset VY to eliminate the scaling set in spm_fmri_spm_ui.m.
Hope this helps!
-Tom
-- Thomas Nichols ---------------------------------------------------------
Department of Biostatistics
School of Public Health phone: +1-734-936-1002
University of Michigan fax: +1-734-763-2215
1420 Washington Heights email: [log in to unmask]
Ann Arbor, MI 48109-2029 web: http://www.sph.umich.edu/~nichols
---------------------------------------------------------------------------
> On Mon, 1 Dec 2003, Thomas E Nichols wrote:
>
> > > In response to my question below, Russ wrote that data are generally
> > > globally scaled to a mean of 100 on the beta0 image I think (the
> > > baseline)? Is this true by default for spm2 data as I found several
> > > subjects with within-brain average beta0's in the 300's across all
> > > voxels?
> >
> > I've observed this too. For an explanation see
> >
> > http://www.jiscmail.ac.uk/cgi-bin/wa.exe?A2=ind0204&L=spm&P=R11588&I=-1
> >
> >
> > For more on how to robustly estimate typical intensity with the mode
> > see Appendix II in
> >
> > Wen-Lin Luo and Thomas E. Nichols,
> > Diagnosis and exploration of massively univariate neuroimaging models.
> > NeuroImage, Volume 19, Issue 3, July 2003, Pages 1014-1032
> >
> > -Tom
> >
> >
> > -- Thomas Nichols -------------------- Department of Biostatistics
> > http://www.sph.umich.edu/~nichols University of Michigan
> > [log in to unmask] 1420 Washington Heights
> > -------------------------------------- Ann Arbor, MI 48109-2029
> >
>
>
-Tom
|