Biomedical Engineering Reference
In-Depth Information
Figure 6.
Top: representative slices from the level of the precentral gyrus, with simulated
atrophy indicated by the arrows (left = before, right = after uniform 30% atrophy within the
gyrus was applied). Bottom: regions detected by the RAVENS analysis, overlaid on the
average WM RAVENS maps of the 24 individuals. The two detected regions were exactly
where atrophy was simulated. Reprinted with permission from [2]. Copyright c
2001,
Academic Press. See attached CD for color version.
sity increases during contraction. Consequently, tissue density in the template's
(stereotaxic) space is directly proportional to the volume of the respective structure
in its original form. Therefore, regional volumetric measurements and compar-
isons are performed viameasurements and comparisons of the respective RAVENS
density maps. One RAVENS map is generated for each tissue of interest, typi-
cally GM, WM, and CSF. In [2] we validated RAVENS on 24 MR images having
synthetic atrophy. Specifically, we randomly selected standard SPGR images of
12 BLSA subjects, and we outlined the precentral and superior temporal gyri in
all of them. We then introduced a uniform 30% volumetric contraction in these
two outlined gyri, thereby generating another 12 images with synthesized atrophy
in them. Figure 6 (top) shows cross-sections of a typical image before and after
contraction of the precentral gyrus (segmented images are shown).
We then used RAVENS to determine the 24 respective brain tissue density
maps, and applied a pointwise statistical analysis to them via paired
t
-tests. Re-
gions of statistically significant differences between the two sets of 12 are shown
in Figure 6 (bottom), overlaid on the averageWMRAVENS map of the 24 subjects
(used for reference). The highlighting of the two regions in which atrophy was
introduced shows the spatial specificity of RAVENS. In [2] we also compared the
