Biomedical Engineering Reference
In-Depth Information
When image volumes are transformed into this space and resampled on the
same voxel grid such that all brains have the same orientation and size,
voxel-by-voxel comparisons across data volumes from different populations
are possible, since each voxel (
) points in
the brain-based coordinate system. The transformation to this coordinate sys-
tem also provides a means for enhancement of functional signals by averag-
ing images in this space.
i
,
j
,
k
) corresponds to the same (
x
,
y
,
z
32
This paradigm allows information (anatomical,
metabolic, electrophysiological, chemical, cytoarchitectonic) from different
brains to be spatially organized and cataloged by mapping all brains into the
same coordinate system.
31
Finally, in the original Talairach spirit, the coordi-
nate corresponding to a particular structure, as defined by an atlas in this
coordinate system, can be used to
its location in a subject's brain volume
when mapped into the same space. The accuracy of prediction depends
greatly on the residual anatomical variability remaining after mapping a
brain into this coordinate system. For thalamus and basal ganglia structures,
this mapping and the corresponding prediction are quite accurate because
anatomical variability of these structures in this frame of reference is low. The
predictive accuracy is lower
predict
for cortical structures, mainly due to the signifi-
cantly higher anatomical variability in this space.
When used for interpretation of cortical regions, there are a number of diffi-
culties associated with the Talairach atlas. (Note the distinction between the ste-
reotaxic
.) The atlas is derived from a single
cadaver brain of a right-handed, 60-year-old European female, and thus may
suffer from postmortem artifacts. The atlas is defined on one hemisphere and
reflected to the other hemisphere, making it completely symmetric and ignor-
ing well known left-right hemispheric differences. With the exception of the
upper midbrain, the atlas excludes the brainstem and cerebellum. It has variable
slice separation, up to 4 mm. While it contains transverse, coronal, and sagittal
slices, it is not contiguous in 3D nor is it entirely consistent among the three rep-
resentations.
atlas
and the stereotaxic
space
33
Finally, and perhaps most importantly, it is derived from a single
subject, and thus it cannot represent statistical models of anatomical variability.
Despite the difficulties associated with the Talairach atlas, there are a num-
ber of significant advantages to using stereotaxic space for brain imaging
studies, the most important of which is that this space provides a conceptual
framework for the completely automated 3D analysis across subjects in cohort
studies. In particular, registration of brain image volumes to this space:
•
Facilitates comparisons 1) across time points for an individual sub-
ject, 2) between subjects, 3) between different groups of subjects,
and 4) across acquisition sites
•
Permits voxel-to-voxel averaging between subjects to detect sub-
tle signals by increasing the signal-to-noise ratio
32
34
•
Allows the use of spatial masks for postprocessing
•
Allows the use of spatial priors (e.g., for classification)
•
Allows the use of anatomical models (e.g., for segmentation)
