|Introduction||Specific Aims||Volumetric Data Sets||3D Bilateral Filtering||EM-Based Classification||Classification and Segmentation||Collaborators|
A major goal in radiation therapy is to deliver a high radiation dose to the perceived tumor volume while minimizing the dose to surrounding uninvolved tissues. Although radiation therapy cures a large fraction of patients treated with this modality, the incidence of local failure remains a problem and the radiation-induced side effects impact the quality of life for many cancer patients. One of the reasons for local failure and increased side effects is the assumption commonly made in radiation therapy that the geometry of the patient’s anatomy is invariant relative to what is obtained at the time of the initial 3D imaging (usually Computed Tomography or CT) performed for treatment designs. With the availability of novel imaging techniques, tumor volume and normal structures can be defined in “4D” images, i.e, sets of 3D images acquired at specified intervals of time, the interval being dependent upon the specific radiotherapy problem. We hypothesize that taking into consideration of these time-dependent changes in patient’s anatomy and tumor volume will improve treatment success.
Specially, we propose to develop non-linear deformable registration techniques to take into account time-dependent changes in shapes, sizes, and locations of anatomical structures in designing and delivering radiation treatments. The resulting process would be called “four-dimensional radiotherapy” or 4DRT.
The proposal is a novel engineering application in that it combines fundamentals of image analysis, biomechanics, shape modeling, radiotherapy and quantitative validation. It provides an engineering solution to an important clinical problem.
Lei Dong, Ph.D.
Assistant professor, Department of Radiation Physics
University of Texas M.D. Anderson Cancer Center