Hi Silke,
you could well first run a standard segmentation/normalisation using SPM5
and its standard priors. The normalisation "bias" as you call it would
arise in a situation with one group "reaching" the standard space and the
other group not fully reaching it. With MCI the expected brain
volume/cortex changes are rather subtle, further supporting to neglect
such potential asymmetry first.
SPM5 will probably manage to warp these rather mild changes to the
standard space. Modulation of the images is recommended of course.
(Producing own priors from a small, typically clinical sample is probably
not improving much and not absolutely necessary. Some posting are online
with the keyword "customised priors" or "templates" & discussion of this)
br,
Philipp
|