Print

Print
Another potential problem with the procedure that Michael described is that the number of activated voxels is an unreliable measure of activation, as shown by Cohen & Dubois:

http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=pubmed&dopt=Abstract&list_uids=10398975&query_hl=3&itool=pubmed_docsum

Fortunately this measure seems to have waned in the literature in the last couple of years.

I agree with Joe that the best approach is to examine your activations using a good atlas and describe the anatomical regions that are activated. If you want to report something quantitative, then one approach would be to use the probabilistic cytoarchitectural maps from the SPM Anatomy Toolbox (http://www.fz-juelich.de/ime/spm_anatomy_toolbox) and compute signal change within a region of particular probability (e.g., the region with >50% likelihood of being BA 45).  These maps take into account the variability that is glossed in the AAL map.  Unfortunately these maps only cover some of the Brodmann's areas.   There are unfortunately no comprehensive probabilistic sulcal/gyral anatomical atlases that are publicly available, to my knowledge

cheers,
russ


On Apr 25, 2006, at 7:41 AM, Joseph Devlin wrote:

How do you get those local maxima? I don't know the appropriate procedure

I think they are reported by default in the feat web output pages but they can also be generated using the cluster command.  If I remember correctly, the flag is --olmax.  It should be clear from help screen if you type cluster without any arguments.

I don't mean to be critical but if I were a reviewer, I wouldn't be entirely happy with the method you describe.  The problem is that the AAL brain regions were determined on the macroanatomy of one brain (Colin27) and therefore are not representative of any other brain, let alone a group sample.  Tzourio-Mazoyer and colleagues acknowledge this explicitly in their paper.  So the number (or percentage) of voxels in each of the AAL regions is not accurate and the precision of the reporting hides this fact.  In addition, if not all AAL regions were imaged (or some where only partially imaged), this too may mislead the reader regarding the the amount of activation in any given region.

Wouldn't it be more appropriate to describe the activations yourself, based on their extent into various macroanatomic regions seen in the group mean structural?

- Joe

---
Russell A. Poldrack, Ph.d.
Associate Professor
UCLA Department of Psychology
Franz Hall, Box 951563
Los Angeles, CA 90095-1563

phone: 310-794-1224
fax: 310-206-5895
email: [log in to unmask]
web: www.poldracklab.org