Publication
A Training Method for Prostate Delineation Based on CT- and MR-Images
Conference Paper/Poster - Mar 31, 2011
Collon Joop, Schiefer Johann, Stolz Anja, Naef Gabriela, Putora Paul Martin, Schneider Markus, Fretz Christian, Plasswilm Ludwig
Units
Contact
Citation
Type
Conference Name
Publisher Proceedings
Publication Date
Pages
Brief description/objective
Objective: The implementation of dynamic IMRT techniques and
improved imaging methods allows a reduction of the high dose area and
improved sparing of normal tissue. Therefore it is of even greater
importance, that the target volume has to be delineated with the highest
possible accuracy. This study presents a training tool for prostate
delineation, based on CT and MRI images, and a mathematical
formalism developed for this purpose.
Materials and Methods: At the beginning of this study, delineation
rules have been compiled and discussed with the involved physicians as
well as an MRI specialist. Five physicians of our radiation oncology
team delineated the prostate; first based on CT, then again on fused CT
and MR images. Applying Boolean methodes, the structures have been
compared with the prostate structure, delineated by two experienced
radiation oncologists. Additionally, the integral coincidence of the
structures delineated by the participants has been evaluated. Ten patients
have been contoured and discussed immediately after evaluation.
Results: The study encouraged the participants to deal with well defined
delineation rules. The largest delineation advances had been observed at
the beginning of the study. But, since the investigated cases diverged in
stage and image quality, a clear numerically proven progress is difficult
to assess on the basis of only ten patients, however a positive trend could
be observed.
Conclusion: Numerically based training tools are suitable to measure the
contouring progress in an objective manner. It is obvious that the
reference structure can diverge from the real prostate structure. This
method can be applied to other sites.