Dear SPM community,
after reposting my questions to the list I have received yet another curious
inquiry, but nothing else. As the problem of evaluating the percent signal
change in a small volume of interest seems to lie in the center of SPM99's
blind spot, I came up with a hack "solution". Basically, I think that the
answer lies in averaging the beta values for the voxels in the small VOI.
These values aren't returned by spm_regions to spm_results_ui, but this can
be easily changed by changing line 543 in spm_results_ui to:
'Callback','[Y,xY,beta]=spm_regions(SPM,VOL,xX,xCon,xSDM,hReg)',...
and line 1 in spm_regions:
function [Y,xY,beta] = spm_regions(SPM,VOL,xX,xCon,xSDM,hReg,xY)
the beta values can be accessed and averaged in the matlab workspace. It is
my understanding that in a globally scaled simple box-car design they will
represent percent change relative to the mean brain signal.
Any comments will be greatly appreciated.
Shy Shoham
>Dear SPM community,
> I have decided to repost my questions since I have received no answer, and
have actually received
> queries from others interested to know the answer (I'm not alone)...
> I think that this is a fairly important capability, and have seen studies
that reported results of
> %rCBF or %change calculated from a small region. How can I do that in
SPM99? I have only been able
> to calculate %change in a single voxel or the full "eigen time-series" in
a V.O.I., but have no idea > how to get the % signal change (a scalar) in a
VOI without going voxel by voxel or regressing the
> time series with my box-car design...
> What am I missing?
> Shy
> Dear SPM gurus,
> I was able to find in previous postings explanation of how to claculate
the (mean brain normalized) percent
> change in a certain voxel (using plot -> contrast of parameter estimates),
and how to obtain the time series > of changes in a volume of interest
(using the new spm_regions function). I was wondering if you could assist >
me with the following two questions:
>
> 1. How can one obtain the percent change estimate for a small V.O.I., i.e.
a single number rather than the
> full time series? I'm getting too much variance when looking at the
percent change in a single voxel, and I > think that for instance, the mean
beta values in a VOI would be a better representative.
> 2. Is there a simple way of finding the location of (global and/or local)
maximal signal change, rather than > the location where the t statistic is
maximised?
> thanks in advance,
>
Shy
|