Biomedical Engineering Reference
In-Depth Information
concluding that there is no point in delineating volumes of interest to
the last millimeter or so. I believe this is a wrong understanding of
the matter. Very often it is not the appreciation of gross tumor
per se
which defines the surface of the GTV and/or CTV. Rather, it is the
judgment that some normal tissue boundary sets a limit to tumor
extension, or defines a volume which is to be excluded from the high
dose region (e.g., the optic chiasm in the treatment of a base-of-skull
sarcoma). This is the case even in Figure 3.19 where the observers
have clearly made the judgment that the CTV does not extend into, or
beyond, the inner table of the skull and in that neighborhood all four
contours are highly congruent. Such constraints on the target volume
due to adjacent normal tissues can often be delineated very accurately
and are often the basis for spatially accurate and tight target volume
delineation.
Delineation of uninvolved normal tissues and organs
The body contains hundreds if not thousands of anatomic structures.
It is, of course, neither practical nor necessary to delineate them all.
Usually the radiation oncologist will identify those structures which
may be important in designing and characterizing the treatment. Even
then, it is not at all uncommon to require in excess of a dozen
structures to be delineated. This means that manual drawing can be
very burdensome and this is the main impetus for the current interest
in finding ways to extract anatomic structures automatically.
One attractive approach to what is called automated “feature
extraction” is the use of a digital atlas of normal anatomy. In this
method, a prototypical patient is created with a large number of
anatomic features already delineated. This atlas must then be “fit” to
the anatomy of the given patient. There is no problem, in principle, in
creating such an atlas. It is time-consuming, of course, but need only
be done once. (Actually, more than once, since one needs different
atlases for males and females and, probably for very differently sized
or developed individuals - fat and thin, child and adult.) There are
two classes of problems encountered in the process of matching an
atlas to a patient. First, a deformable registration will certainly be
required, with the associated problem of how to match the atlas
information to the information contained in the images. Second,
the normal anatomy is very likely to be distorted by the tumor. The
tumor may replace parts of the normal anatomy, and may displace
and distort normal anatomy. These effects will, of course, not be
