Dear Matthew,
> Our group plans on combining data from two different PET scanners in
> our OCD/DEP study analysis. One of our scanners is Siemens neuro ECAT
> 831 and the other is Siemens ECAT EXACT HR+961.
>
> Recently, I smoothed all our images to 16 mm, and we have tried
> combining the data in various SPM analyses.
> However, some dorsal white matter regions seem to always show
> activation. However, this seems to only occur in the combined
> analyses. Doing two separate analyses for each data set does not
> produce these results.
>
> We would like to know whether this procedure is valid. Further, are
> there any design modifications we need to make to resolve this issue?
Clearly combining data from two different scanners is not ideal. If
find yourself in a position where you cannot avoid doing this then you
could model scanner-specific effects and 'effect of interest' x scanner
interactions. This is implicit if you enter the scanner-specific data
as two studies and make inferences using contrasts that span both.
White matter activations are sometimes seen when the statistical model
for the confounding effects of global activity is improperly
specified. This can occur when global activity, from two scanners,
differs either in terms of the means or variances. If this is the case
then a scaling normalization may be more robust.
I hope this helps - Karl
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|