Skip to main content
eScholarship
Open Access Publications from the University of California

Computational and Physical Quality Assurance Tools for Radiotherapy

  • Author(s): Graves, Yan Jiang
  • et al.
Abstract

Radiation therapy aims at delivering a prescribed amount of radiation dose to cancerous targets while sparing dose to normal organs. Treatment planning and delivery in modern radiotherapy are highly complex. To ensure the accuracy of the delivered dose to a patient, a quality assurance (QA) procedure is needed before the actual treatment delivery. This dissertation aims at developing computational and physical tools to facilitate the QA process. In Chapter 2, we have developed a fast and accurate computational QA tool using a graphics processing unit based Monte Carlo (MC) dose engine. This QA tool aims at identifying any errors in the treatment planning stage and machine delivery process by comparing three dose distributions: planned dose computed by a treatment planning system, planned dose and delivered dose reconstructed using the MC method. Within this tool, several modules have been built. (1) A denoising algorithm to smooth the MC calculated dose. We have also investigated the effects of statistical uncertainty in MC simulations on a commonly used dose comparison metric. (2) A linear accelerator source model with a semi-automatic commissioning process. (3) A fluence generation module. With all these modules, a web application for this QA tool with a user friendly interface has been developed to provide users with easy access to our tool, facilitating its clinical utilizations. Even after an initial treatment plan fulfills the QA requirements, a patient may experience inter-fractional anatomy variations, which compromise the initial plan optimality. To resolve this issue, adaptive radiotherapy (ART) has been proposed, where treatment plan is redesigned based on most recent patient anatomy. In Chapter 3, we have constructed a physical deformable head and neck (HN) phantom with in- vivo dosimetry capability. This phantom resembles HN patient geometry and simulates tumor shrinkage with a high level of realism. The ground truth deformation field can be measured from built-in surface markers, which is then used to verify the accuracy of an important ART step of deformable image registration. Our experiments also demonstrate the feasibility of using this phantom as an end-to-end ART QA phantom with an emphasis on testing the dose deliver accuracy

Main Content
Current View