Author ORCID Identifier


Date of Graduation


Document Type

Dissertation (PhD)

Program Affiliation

Medical Physics

Degree Name

Doctor of Philosophy (PhD)

Advisor/Committee Chair

Laurence E. Court, Ph.D.

Committee Member

Carlos E. Cardenas, Ph.D.

Committee Member

Clifton David Fuller, M.D., Ph.D.

Committee Member

Rebecca M. Howell, Ph.D.

Committee Member

Anuja Jhingran, M.D.

Committee Member

Tze Yee Lim, Ph.D.

Committee Member

Christine B. Peterson, Ph.D.


Creating a patient-specific radiation treatment plan is a time-consuming and operator-dependent manual process. The treatment planner adjusts the planning parameters in a trial-and-error fashion in an effort to balance the competing clinical objectives of tumor coverage and normal tissue sparing. Often, a plan is selected because it meets basic organ at risk dose thresholds for severe toxicity; however, it is evident that a plan with a decreased risk of normal tissue complication probability could be achieved. This discrepancy between “acceptable” and “best possible” plan is magnified if either the physician or treatment planner lacks focal expertise in the disease site.

Many clinics implement expert-peer review programs, where each treatment plan is reviewed by other radiation oncologists with the same disease specialization. These expert peer review programs are not able to be implemented at small clinics, which represent the majority of clinics around the world, due to limited staff and resources. Consequently, a scalable peer review approach to ensure that patients receive high-quality radiation plans is an unmet clinical need for many centers.

The purpose of this study was to develop automated treatment plan quality assurance tools that can provide expert peer review, without the need for actual access to teams of specialized radiation oncologists. To accomplish this, we trained deep learning models to predict patient-specific optimum-achievable 3D dose distributions for radiotherapy plans of head and neck cancer patients. We conducted experiments by varying the deep learning architectures, loss functions, data augmentation techniques, and CT normalization methods to determine the top-performing model. We then tested the application of dose prediction to automatically identify suboptimal head and neck plans and benchmarked its performance against manual physician review. Finally, we tested the translatability of our approach for plan quality assessment to another disease site—gynecologic cancers. We trained a deep learning model to predict high-quality dose distributions for VMAT plans for patients with gynecologic cancers and tested the usability of predicted dose distributions to help improve plan quality by guiding plan re-optimization. The tools developed in this work are expected to be integrated into a web-based automated, expert, peer review system, enabling clinics around the world to receive treatment plan recommendations of the same quality as those offered by expert, specialized radiation oncologists. The implementation of an automated, expert peer review system will address current disparities in expertise and resources of clinics around the world.


artificial intelligence, deep learning, quality assurance, peer review, dose prediction

Available for download on Thursday, April 25, 2024