Morphometric Data - An Invariant Approach to Statistical Analysis of Shapes - page 35

Landmarks from CT images
CT provides a two-dimensional representation of an anatomical sec-
tion through the biological object. A useful metaphor is a loaf of bread
that has been sliced. The pieces of bread are numbered consecutively
starting at one end so that the relationship of one slice to the others
and to the whole is known. You can remove a slice and see what the
inside of the bread looks like anywhere along the loaf and replace the
slice to reconstitute (reconstruct) the loaf.
When a biological form undergoes CT scanning, images are made
that represent slices through the form. This is done by means of nar-
row collimated x-ray beams that penetrate the form and produce x-ray
transmission measures. Large numbers of these measures are taken
from many more beams targeted through the same cross section from
many (approximately 200-1200) directions, some overlapping. The
totality of these beam measurements are recorded and combined to
reconstruct a two-dimensional matrix of coefficients that represent the
linear x-ray attenuation coefficients of the scanned slice (Zonneveld,
1987). Each element of this matrix is referred to as a pixel, and each
pixel has a known size that is used to produce a scale for the image.
The reconstruction of all pixels produces a picture of the entire slice
based on the linear attenuation coefficients that follow a scale having
air at one extreme and a dense substance (e.g., mercury) at the other
extreme ( Figure 2.6 ). All tissues are assigned a score based on this
scale.
Two-dimensional landmark coordinate data can be collected direct-
ly from slice images following methods similar to those discussed
above for serially sectioned specimens. As long as the scale (pixel size)
is known, 2D landmark locations can be recorded according to pixel col-
umn and row and later converted into metric units. In order to obtain
a three-dimensional coordinate location for a landmark on a slice,
information about the location of the point in the plane of the slice (X,
Y coordinates), the thickness of the slice, and the location of that slice
within the whole (Z coordinate) must be known. First, the slice con-
taining a landmark of interest is identified. Next the X, Y coordinate of
the landmark is recorded according to pixel column and row. The Z
coordinate for each landmark is assigned according to table position
(the position of the table for that particular slice as it moves through
the x-ray tube), or it can be assigned according to a sequential num-
bering system from one end of the specimen to the other. The thickness
of the slice must be accounted for in this numbering system. When