One thing worth bearing in mind is that registration is not simply a
topology problem -- at the micro-structure level brains simply are not the
same across individuals. This is no doubt due to genetic, developmental,
and morphological issues but it means that there is *no* registration,
linear or not, which will "match" one brain to another. In addition, this
means there is no meaningful "canonical" brain -- it is a completely
arbitrary choice if one brain is chosen as a template.
Having said that, registration is still a very important process because in
most functional imaging studies we collect data from more than one
individual because we want to draw inferences at the population level. In
order to do that, we need to align these images in a reasonable fashion for
combining the results in order to do our statistics. Given the inherent
variability across subjects, the most reasonable template is a sufficiently
large average of brains such as the MNI 152 mean brain because we assume
that this is representative of the population.
Although some groups refer to images aligned to this template as "Talairach
space", they typically mean that they are using the coordinate conventions
proposed by Talairach and Tournoux -- namely an origin in the medial AC
with x-, y-, and z- coordinates. They do not mean that the coordinates
themselves correspond to those in the T&T atlas. Obviously this
terminology is somewhat confusing and consequently many people have changed
to saying that their images were registered into "standard space" or to the
MNI template." To the best of my knowledge, all of the major software
packages use the MNI152 template for registration by default and none use
the T&T brain (I think SPM95 was the last time that was used by
default). Consequently, using the T&T atlas as a guide to the anatomical
location of activations is not a good idea. In contrast, assuming that
structural images were also collected, then the relevant macroanatomical
information is available to describe the activations in terms of the real
anatomy, as Christopher suggested.
Finally, Cinly's point regarding the many sources of uncertainty in
localisation -- imperfect registration, pre-registered morphological
differences in both the structural and functional anatomy, biases towards
draining veins -- is well taken. I think one would be hard pressed to
argue that a peak 3mm from another in a different study was truly different
and 6mm isn't much better. Ideally the peaks could be compared based on
their anatomical positions relative to macroanatomic landmarks (eg sulci)
and if there is a systematic difference then that may be informative (e.g.
on the crest of a gyrus vs in a sulcus).
Joseph T. Devlin, Ph. D.
FMRIB Centre, Dept. of Clinical Neurology
University of Oxford
John Radcliffe Hospital
Headley Way, Headington
Oxford OX3 9DU
Phone: 01865 222 494
Email: [log in to unmask]