Hi Max
> Dear SPMers,
>
> We have had some lively debate about the normalization quality for
> several brains with large ventricles (see here
> https://docs.google.com/fileview?id=0B2N0iTxTkqzWYjMzMzQ1YmQtYzk0Yi00MTE3LWI4NjUtY2JhMmM2OTFlNWEz&hl=en_GB
> <https://docs.google.com/fileview?id=0B2N0iTxTkqzWYjMzMzQ1YmQtYzk0Yi00MTE3LWI4NjUtY2JhMmM2OTFlNWEz&hl=en_GB>
> ).
>
> 1) Is there is any objective, ideally quantitative, way to tell
> whether the normalization result is high enough quality to use in a
> group analysis?
well I don't know about 'good' but here is what I've done to make my
reviewers happy
- use the VBM toolbox (http://dbm.neuro.uni-jena.de/vbm/vbm5-for-spm5/)
to look at how my group(s) is (are) homogeneous computing a distance
from the normalized brain to the mean
- do this analysis running t-tests between groups on the distance to the
mean and to the template (for this I made some minor modif to the
toolbox that you can download @
http://www.sbirc.ed.ac.uk/cyril/cp_download.html)
I also displayed averaged normalized brains, worse and best segmented
brains as supp material - this was enough to convince people that there
was no normalization issues .. you can find an example of that @
http://www.sbirc.ed.ac.uk/cyril/cp_dyslexia_research.html and the
related article HBM 2009 30(7)
best
cyril
--
The University of Edinburgh is a charitable body, registered in
Scotland, with registration number SC005336.
|