Biomedical Engineering Reference
In-Depth Information
have been particularly important in obstetrics, cardiology, and urology. Although
3DUS has been shown to be useful and is used clinically in a variety of applications,
progress is still required in various aspects of this imagingmodality for it to achieve
its full potential. In this chapter we have examined one application of 3Dultrasound
that requires fast, accurate, precise, and robust segmentation of the prostate for use
in prostate brachytherapy.
As discussed above, a DDC-based prostate segmentation tool may allow the
development of an intraoperative prostate therapy technique in which all steps are
carried out at one session. Some of the necessary aspects of the complete proce-
dure have already been developed and have been reported elsewhere. A 3D US
image of the prostate can be obtained quickly and a dose plan can be calculated
with that image. The location of the implantation needles used to implant the
radioactive seeds can be accurately localized in 3D and can be verified using fast
3DUS imaging tools. However, for the procedure to be effective and accurate, fast
segmentation of the prostate is required for all aspects of the procedure, includ-
ing dose planning prior to radioactive seed implantation, re-planning during the
procedure due to changes in the prostate, and post-implantation dose calculation
to determine if any regions of the prostate are underdosed. The DDC-based seg-
mentation tool described in this chapter has been demonstrated to be an excellent
candidate to achieve these goals. We have shown that the segmentation in 3D can
be achieved within seconds, and the results are accurate and precise. Nontheless,
some improvements are still necessary to make this approach better.
DDC-based segmentation requires good initialization of the prostate to con-
verge to the desired boundary. We described a method of initialization requiring
the user to input four or more initialization points. However, to minimize the work-
load on the user, interactive initialization should require minimal input or perhaps
no input at all. Most prostate shapes are approximately ellipsoidal; however, devi-
ations from a simple shape exist, especially in diseased prostates. To handle these
difficult cases, our approach has been to provide the user with boundary editing
capabilities for interactive editing of the boundary. However, manual editing may
result in inaccuracy and variability and can be time consuming if many editing op-
erations are required. Since the difficult cases occur often enough, it is important
to develop alternate initialization approaches that require no user interactions.
Even with automated and accurate initialization, boundary editing facilities
are necessary since boundary segmentation errors may occur because the prostate
boundary is either missing or very weak. Although the current editing tools are
effective, they require the user to employ a mouse and select the erroneous points
on the computer screen for editing. This operation has been found to be difficult in
a busy and sometimes crowded operating room, where the physician may have dif-
ficulty in manipulating a computer mouse to edit points a screen. Thus, improved
approaches for editing would be of great benefit.
Since segmentation tools are to be used clinically, validation of the segmen-
tation is critical. While a great deal of effort has been spent on the development of
