Biomedical Engineering Reference
In-Depth Information
during treatment simulation. The contours of the proximal seminal
vesicle (part of the target volume)
is shown in orange; the bladder is
shown in blue; the rectum is
shown in green; and the femoral
heads are shown in purple.
Figure 3.18b shows a CT image of
(a)
(a)
the same patient during one of
the treatment sessions. The CT
(b)
(b)
image was acquired using an
in-treatment-room CT-on-rails. It
can be seen that the rectal gas and
bladder filling have changed the
anatomy. The contours overlaid on
the CT image after a rigid-body
image registration of the pelvic
bones could not match the patient's
(c)
(c)
anatomy. Figure 3.18c shows
the same set of contours after
performing a deformable image
registration. It is clear to the eye
that these contours match the
anatomy much better than in
Figure 3.18b. This particular auto-
Figure 3.18. Comparison of
rigid and deformable registration
of contours (see text). Figure
courtesy of
Lei Dong
, MD
Anderson Cancer Center, USA.
matic deformable image registra-
tion technique was developed by
Wang
et al.
(2005).
The use of diagnostic imaging studies
Diagnostic studies are usually done: with the patient not in the
treatment position; with only a partial field of view; and with possibly
spatially distorted images. The emphasis on treatment accuracy has
given rise to a mistaken impression that diagnostic studies may not be
useful in designing the geometric aspects of a treatment. However,
geometrically inaccurate studies often contain a wealth of valuable
information which can be transferred to the planning CT. This might
be done by deformable registration. But, it may often be done by
visual inspection, using simple common sense. If, say, a diagnostic
study shows a tumor fixed to C3 and extending down to mid-C4, up
to the top of C2, and anteriorly by 3 cm, then this information can be
manually transferred to the planning CT accurately and with ease.
