Biomedical Engineering Reference
In-Depth Information
(PET) or functional magnetic resonance imaging (fMRI) to correlate
structure and function in normal and diseased subjects. The combination
of
imaging data with interactive computerized visualization tools
has significantly improved the understanding of gross anatomy in the
living brain. Even with these new tools, traditional technical and practi-
cal difficulties have conditioned many neuroscientific studies to extrapo-
late findings from a relatively small sample of the population. In this chapter,
our research program at the Brain Imaging Centre (BIC) of the Montreal
Neurological Institute (MNI) will be used as an exemplar of the work in
this field and will illustrate our perspective on this fast moving area of
research.
A significant part of our research program has concentrated on the devel-
opment of computerized algorithms for automated analysis of large ensem-
bles of anatomical and functional data. These data arise from cohort studies
where the number of subjects is counted in the hundreds, if not thousands.
These tools are the result of our participation in the International Consor-
tium for Brain Mapping (ICBM
in vivo
1
) where the goal has been to address certain
weaknesses associated with classical brain atlases (or atlantes). Such atlases
are derived from a single brain, or brains from a small number of subjects.
2-6
Even though the brains selected for use in this atlas may be considered nor-
mal, they may still represent an extreme of the normal distribution. Com-
parison of a given subject's brain is achieved by aligning the brain with the
atlas, and then using simple scaling factors in an attempt to account for
variability in brain size. Unfortunately, the limited number of examples
used to create the atlas prohibits quantification or representation of normal
anatomical morphometric or functional variability. Thus, it is difficult to
judge the degree of normalcy of a brain under study by comparison with
one of these atlases.
In keeping with the goals of the Human Brain Project,
7
the thrust of the
ICBM project is to characterize normal anatomy and function by building
normative databases and the tools required for interaction with these data-
bases. The primary aim of the first phase of the ICBM project was to quan-
tify the normal range of morphometric variability and to build a probabilistic
atlas of structural anatomy. In the second phase, similar tools are now
being developed to capture and characterize variability of functional
regions.
The methods developed in the context of the ICBM project are directly
applicable to cohort studies where the goal may be to characterize the brain
anatomy of a specific group of subjects or to quantify differences between
two groups. For example, one might be interested in using subtle anatomi-
cal differences, such as differential rates of cerebral atrophy, to discriminate
between groups of patients treated with different medications. For many of
these studies, data such as blood volume, blood flow, glucose utilization, or
tracer uptake must be estimated for a particular region of interest from the data
set corresponding to each subject. Traditional manual definition of such regions
is difficult, time consuming, and potentially unsuitable if inter- and intraobserver
