These are good questions, but difficult to answer, so thoughts from others would
be very welcome.

I am not sure that the problems are completely insurmountable, but it may be worth
treating the results of the analyses with a little bit of healthy scepticism.
Systematic shape differences between the groups of subjects may bias your results
slightly.

I guess the general question you are trying to answer is something like: "does this
group activate more than the other".  The answer to this one depends on how you
quantitate the 'amount of activation' in different brains.  For example, it could be
in units of blood flow per unit of whole image volume, or blood flow per unit of grey
matter, or per unit of hippocampal volume.  Doing the stats with all these different
sets of units will give you subtly different results.

In addition, the images are warped during spatial normalisation, which leads to volume
changes in the different brain structures.  Because the spatial normalisation only
uses about 1000 parameters, it only models smooth deformations.  Because of this, the
size of the hippocampus after spatial normalisation will be based on the region all around
the hippocampus, and not just the hippocampus itself.  Therefore, all the hippocampi
are unlikely to be exactly the same size after spatial normalisation.

If you wish to include hippocampal volume as a confounding effect in the model, then it
would probably be better to determine its volume after spatial normalisation rather than
before.  However, your activations may be from only one part of the hippocampus, whos
volume may be completely unrelated to the whole hippocampal volume. 

Good luck,
-John



| In the vein of the present interest in templates, normalization and elderly subjects...
| 
| We are interested in comparing two groups of elderly subjects (cognitively impaired/ 
| not cognitively impaired).  One of the hurdles we are are trying to overcome or to consider 
| concerns differences in patterns of atrophy for these two groups. Although both groups "suffer" 
| from a general  brain atrophy, the cognitively impaired group is known for having a 
| disproportionately large hippocampal atrophy (in comparison to the rest of their atrophy) . 
| 
| 1) From a "spatial normalization" perspective, if one desires to conduct a direct group 
| comparison, is this an insurmountable problem?
| 
| 2) Assuming normalization is not a problem, what are your thoughts on using hippocampal 
| volume as a covariate in the analyses?  In other words, can adjusting for hippocampal 
discrepancies
| through spatial normalization be viewed as taking care implicitly of hippocampal size, rendering 
the use 
| of covariates obsolete and actually redundant? 

  

| I have realized that question no.2 may have not been very clear.  Here's an attempt
| at clarifying the issue... 
| 
| 2) Assuming normalization is not a problem, what are your thoughts on using hippocampal 
| volume as a covariate in the analyses?  Can adjusting for hippocampal discrepancies
| through spatial normalization be viewed as taking care implicitly of hippocampal size, rendering 
the use 
| of a covariate (hipp. volume) obsolete and actually redundant IF ONE WANTS TO SPECIFICALLY 
| EXAMINE HIPPOCAMPAL ACTIVATION? 
| 



%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%