Hello Silke,
yes, SPM5 allows you do test such hypotheses, basically by two different ways:
1) If you hypothesize that those regions that show atrophy at the group
level are the ones that also predict cognitive decline you could restrict
your analysis to the regions detected before in the group comparison.
(a) E. g. extract the raw volumes from the clusters there, and do some
offline analyses with these values such as partial correlations with your
performance variable.
(b) Alternatively, to screen for a correlation in the detected areas you
can set the cursor to your peak voxel of your current analysis and use the
plot function to create the scatterplot between (adjusted) volumes and the
cognitive score. However, this plot would be across both groups (see below).
2) The areas "atrophied" at the group level need not necessarily correspond
to the areas that predict performance decline best. So sometimes it is
better to do an independent analysis in SPM (multiple regression model)
with your performance variable and other covariates that you want to
correct for. As covariates (of no interest) one would rather take the same
ones you used in your group comparison. You can then ask for a positive or
negative correlation with your test variable by using a simple t-contrast
on the respective regressor.
For both (1) and (2) it is a bit of a conceptual debate if to use both
groups of only the MCI group:
If you pool the groups, in both approaches you can produce artificial
correlations that are more based on the fact that these groups are
"morphologically" and "cognitively" apart. So more "independent" evidence
is provided if the correlation only exists in the MCI group: this would
tell you that the areas are abnormal AND functionally relevant (esp. if
analysis (2) would give you the same areas as your group analysis).
best regards,
Philipp
Max Planck Institute of Psychiatry
NMR Research Group
Kraepelinstr. 2-10
80804 Munich
Mail: [log in to unmask]
Phone: 0049-89-30622-413
|