I am relatively new to fMRI and the mailing list and my question may be
rather naive. I was hoping to do the same as Eli - map the coordinates of
the standard fMRI space (eg Telairach) to regions of the structural brain
atlas. If I understand your point, functional maps do not map directly to
structural maps as everyone is different. Instead a probabilistic approach
can be used. However, how does this relate to differences between
individuals in their functional maps. If there was no difference, then
couldn't an atlas be redefined functionally? If this still remains a
probabilistic map because of inter-subject variability then isn't this a
problem for the smallest region we can identify - a voxel. If this is the
case, does this create issues in interpreting voxel by voxel analyses? This
seems quite fundamental, but probably just reflects my lack of experience in
the field.
Trevor
-----Original Message-----
From: SPM (Statistical Parametric Mapping) [mailto:[log in to unmask]]On
Behalf Of Russ Poldrack
Sent: 04 March 2005 20:58
To: [log in to unmask]
Subject: Re: [SPM] Talairach database
Eli - this is a FAQ and I feel like a broken record, but here goes:
The idea of an automated one-to-one mapping between stereotactic
locations and anatomical structures is highly problematic. The fact
that the Talairach atlas (and sites like the Talairach Daemon) support
this function does not make it any less problematic. There are a
couple of problems. One is that volumetric spatial normalization does
not absolutely align structural features; that is, if you normalize all
of your subjects to a common space, and then classify which particular
brain bits fall at a particular location for each subject, you will see
that there is not perfect alignment across subjects. Various people
(e.g., the MNI group) have produced probabilistic atlases showing this,
and some of this information is available through the Talairach Daemon.
Thus, the best you can do is say that you are in a particular
structure with some probability.
Second, there is the issue of structure-function association. Work by
Amunts and a number of others has shown that cytoarchitecture (e.g.,
Brodmann's areas) does not follow gyral anatomy. This is particularly
the case in places like prefrontal cortex (e.g., see Amunts' work on
area 44/45). Thus, you can't simply go from a stereotactic location to
a Brodmann's area.
Third, the Talairach atlas presents what are essentially guesstimates
about where particular Brodmann's areas fall - they are not based on
direct histology of the Talairach brain. Going from Talairach
coordinates to Brodmann's areas by means of the Talairach atlas is thus
highly suspect.
I will not pretend to have the right answer to this question, but the
approach that I try to instill into my students is to actually
understand the anatomy where they are seeing activation, looking both
at group maps and individuals. This is best done by working with a
high-quality anatomical atlas; we like the Duvernoy atlas. In the end
the goal should be to have a good enough 3-D mental model of brain
anatomy that you can do this without the atlas, but that takes a long
time indeed.
If you feel really strongly about localizing to Brodmann's areas, then
one approach is to use the probabilistic maps created by the Julich
group (http://www.fz-juelich.de/ime/ProbabilityMaps_eng.html). With
these maps, you can provide a probability that your activation is in
any of the particular areas for which there are maps. Unfortunately
these are only available for a limited number of areas.
I recommend that you read Matthew Brett's paper on localization in
functional imaging, which discusses these and other issues quite
nicely.
cheers,
russ
On Mar 4, 2005, at 11:55 AM, Eli Packer wrote:
> Hi,
> I need a way to translate voxel coordinates to their corresponding
> regions. This is the task of Talairach. I saw that there are applets
> that can be downloded but it will not help me since I want to use this
> database in my own code. I also do not want to be involved in
> converting SPM code or such. Just a simple database with pairs of
> coordinate and regions is enough. Any ideas how to obtain it?
>
> Thanks a lot,
> Eli Packer
>
> __________________________________________________
> Do You Yahoo!?
> Tired of spam? Yahoo! Mail has the best spam protection around
> http://mail.yahoo.com
>
---
Russell A. Poldrack, Ph.D.
Assistant Professor of Psychology, UCLA
Franz Hall, Box 951563
Los Angeles, CA 90095-1563
email: [log in to unmask]
phone: 310.794.1224
fax: 310.206.5895
web: http://www.poldracklab.org/
|