Award Date

5-15-2026

Degree Type

Doctoral Project

Degree Name

Doctor of Medical Physics (DMP)

Department

Health Physics and Diagnostic Sciences

First Committee Member

Steen Madsen

Second Committee Member

Cephas Mubata

Third Committee Member

Yu Kuang

Fourth Committee Member

Ryan Hecox

Fifth Committee Member

Brach Poston

Number of Pages

284

Abstract

The increasing complexity of modern radiotherapy demands planning workflows that are efficient, standardized, and dosimetrically robust across diverse disease sites. Knowledge-based planning (KBP) systems such as RapidPlan offer a data-driven approach to automate and improve treatment planning by learning geometric–dosimetric relationships from high-quality clinical plans. In this work, I am evaluating the performance, generalizability, and clinical applicability of vendor-provided RapidPlan models across seven anatomical sites: intracranial SRS, prostate SBRT, right and left lung SBRT, liver SBRT, head and neck, and glioblastoma. Subsequently, a complementary institution-specific SRS model tailored to single-isocenter multitarget workflows was created. Seventy retrospectively selected patients were replanned using RapidPlan-generated objectives under identical beam configurations, prescriptions, and normalization criteria, followed by detailed comparison of target coverage, organ-at-risk sparing, and plan quality metrics using DVH-based and statistical analyses. Across all sites, model-generated plans were compared to manual plans, demonstrating consistent target coverage, reduced hotspot variability, and equal or superior OAR protection. Stress-testing under challenging conditions including metal artifacts and altered VMAT arc arrangements confirmed stable model behavior. The custom “Tobi SRS Model_1,” trained on 77 institutional multitarget cases, further demonstrated strong predictive performance for complex small-target geometries and maintained high-quality dosimetry across multi-lesion scenarios. Patient-specific QA using EPID and high-resolution SRS MapChecK verified deliverability under clinical tolerances. Collectively, these findings establish that using vendor provided models training on clinical data from outside institution can streamline planning, reduce inter-planner variability, and support safe, reproducible radiotherapy delivery, while also highlighting the value of customized models that reflect local planning conventions and clinical priorities.

Keywords

Automated planning; Knowledge based planning; Medical Physics; Radiation Oncology; RapidPlan; Therapeutic Medical Physics

Disciplines

Investigative Techniques | Nuclear | Physics

File Format

PDF

File Size

11000 KB

Degree Grantor

University of Nevada, Las Vegas

Language

English

Rights

IN COPYRIGHT. For more information about this rights statement, please visit http://rightsstatements.org/vocab/InC/1.0/

Download


Share

COinS