Optimization Problems in Radiation Therapy Treatment Planning.

Abstract

Radiation therapy is one of the most common methods used to treat many types of cancer. External beam radiation therapy and the models associated with developing a treatment plan for a patient are studied. External beams of radiation are used to deliver a highly complex so-called dose distribution to a patient that is designed to kill the cancer cells while sparing healthy organs and normal tissue. Treatment planning models and optimization are used to determine the delivery machine instructions necessary to produce a desirable dose distribution. These instructions make up a treatment plan. This thesis studies four problems in radiation therapy treatment plan optimization.

First, treatment planners generate a plan with a number of competing treatment plan criteria. The relationship between criteria is not known a priori. A methodology is developed for physicians and treatment planners to efficiently navigate a clinically relevant region of the Pareto frontier generated by trading off these different criteria in an informed way. Second, the machine instructions for intensity modulated radiation therapy, a common treatment modality, consist of the locations of the external beams and the non-uniform intensity profiles delivered from each of these locations. These decisions are traditionally made with separate, sequential models. These decisions are integrated into a single model and propose a heuristic solution methodology. Third, volumetric modulated arc therapy (VMAT), a treatment modality where the beam travels in a coplanar arc around the patient while continuously delivering radiation, is a popular topic among optimizers studying treatment planning due to the difficult nature of the problem and the lack of a universally accepted treatment planning method. While current solution methodologies assume a predetermined coplanar path around the patient, that assumption is relaxed and the generation of a non-coplanar path is integrated into a VMAT planning algorithm. Fourth, not all patient information is available when developing a treatment plan pre-treatment. Some information, like a patient's sensitivity to radiation, can be realized during treatment through physiological tests. Methodologies of pre-treatment planning considering adaptation to new information are studied.