Biomedical Engineering Reference
In-Depth Information
Figure 8.
Automated segmentation results using 3D HAMMER for subject AD, years
3 and 4 (see Figure 15, top left). The 5% difference in volumes, in this case, is not
readily appreciated visually from these images. (The sections are only approximately
corresponding, since the scans were at slightly different orientations. 3D renderings are
shown on the right.) Reprinted with permission from [3]. Copyright
c
2001, Society for
Neuroscience. See attached CD for color version.
to yield both stable and accurate longitudinal measurements, compared to 3D
warping.
3.5. Diagnosis: Putting It All together
The voxel-based morphometric analysis methods described above have en-
joyed widespread acceptance in the past decade, since they do not rely on any a
priori hypotheses regarding the structures to be measured, but rather apply unbi-
ased analyses of the entire set of data on a voxel-by-voxel basis. Accordingly, they
highlight regions in which there is a statistically significant difference between two
groups, for example. However, the existence of significant differences in certain
brain regions does not necessarily imply that volumetric measurements of those
regions are sufficient to diagnose disease. For example, say that normal control
older subjects differ from patients developing mild cognitive impairment (MCI)
in the volumes of the hippocampus and the entorhinal cortex (ERC), but volumes
of normal and MCI individuals are highly overlapping. In this case, diagnosis
based solely on volumes of the hippocampus and the ERC could be unreliable. In
recent years, interest in integrating voxel-wise mophometric measurements into
tools that can be used for diagnosis has gained interest [31,72,73]. One of the mo-
tivating factors behind these developments is the complex and spatiotemporally
distributed nature of the changes that most diseases cause, particularly in the brain.
For example, the anatomical structures that carry most discriminative power are
likely to depend on the stage of the disease, as the disease progressively spreads
throughout various brain [74], but also on age and other demographic and genetic
[75], since disease is to be distinguished from complex and progressively changing
background normal variations in anatomy and function that may depend on demo-
graphic and/or genetic background. Moreover, disease might cause changes of the
image characteristics beyond those measured by volumetrics, such as brightening
