Actually the thresholds for the masking are set in spm_fmri_spm_ui.m
around the line where is says "Compute Global Covariate." For fMRI an
average is computed for all the voxels in a slice and for voxels to
survive they have to have an intensity that is 100% of that average
or greater. In usual fmri data since you have a whole bunch of 0's
outside the brain (or values close to 0 anyway) it's no problem for
the in-brain voxels to exceed that average. For your data in which
the slices are simulated it's possible that the average (consisting
of noise and signal voxels) has a greater value than your signal
voxels.
You could try changing the value of GM to 0. I don't know what line
it is but it's in spm_fmri_spm_ui.m and the line above it looks like
this.
%-Compute Global variate
%-----------------------------------------------------------------------
GM = 100;
I think that will work, but hopefully John will respond if I'm not correct.
regards,
Darren
THanks Darren. But unfortunately it didn't work. We still get the same
message, i.e.
-----------------------------------------------------------------------
Warning: No inmask voxels - empty analysis!
> In /home/wilkin/spm_spm.m at line 1076
In /software/spm99/spm_fmri_spm_ui.m at line 619
Design parameters : ...intrinsic autocorrelation
Warning: Divide by zero.
-----------------------------------------------------------------------
We also checked to see how many voxels were below the slice mean, and there
are very few ones. So, the error should not be due to few voxels =or> slice
average. Any other suggestions?
thanks
Valeria
Wilkin
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|