Dear SPMers
I've been looking at studies which have investigated correlations
between brain volume and cognitive covariates in disease groups using
VBM. It seems to me that people are modelling this in at least 3 main
ways:
1) Split patients into 2 groups on the basis of their cognitive score
(a "good" group and a "bad" group), compare each group with its own
matched control group, and present the SPMs side by side. Repeat for
other cognitive variables.
This seems unsatisfactory to me because a) you don't directly
correlate the score with volume and b) more importantly you don't
directly compare the two patient groups, and the visual comparison of
the 2 SPMs is not a statistical comparison.
2) Enter the cognitive score as a regressor in a single model, and
test whether the slope is significantly different to zero at each
voxel. Repeat for other cognitive variables.
This seems to be better in that you can interpret a single SPM as
showing regions in which score is associated with volume, but if you
have more than one variable and they are all modelled separately you
can't really interpret one in the light of the others, although I
think this is what people tend to do; again you would just be
comparing a set of SPMs visually rather than comparing the results
statistically at each voxel.
3) Enter e.g. two cognitive scores as regressors in a single model,
and test whether either slope is significantly different to zero at
each voxel.
This is asking a different question but the interpretation seems
limited: you can show regions in which score A correlates with volume,
having adjusted for score B (and vice versa), but you cannot conclude
that A is associated with those regions significantly more than B,
although I think this is also what people tend to assume.
I'd be very interested to hear how others have approached this
problem, or what people think of my interpretations of the methods
I've summarised above. I don't claim to have done an exhaustive
search of the literature so perhaps there are better/more easily
interpretable solutions out there . . . what is the best way to test
which voxels are associated with test A significantly more than test
B?
Thanks,
Susie
Susie Henley
Dementia Research Centre
UCL, Institute of Neurology
London
|