Author ORCID Identifier
Date Available
4-7-2023
Year of Publication
2023
Degree Name
Doctor of Philosophy (PhD)
Document Type
Doctoral Dissertation
College
Medicine
Department/School/Program
Radiation Science
First Advisor
Dr. Quan Chen
Abstract
Radiotherapy is a frequently used therapeutic modality for cancer patients. Accurately contouring of tumors and organs at risk (OARs) is critical for developing optimal treatment plans in radiotherapy, especially after the implementation of Intensity-modulated radiation therapy (IMRT) and Stereotactic Body Radiation Therapy (SBRT). The manual contouring process is time-consuming and suffers from inter-observer variations. However, manual contouring is often hindered by laborious clinical duties, leading to reduced effectiveness, and increased segmentation errors due to fatigue. Additionally, online adaptive radiation therapy(ART), which has been shown to benefit patient outcomes, places higher demands on contouring and quality assurance (QA) speed.
Recently, deep learning auto-segmentation (DLAS) has emerged as an accurate tool for contouring in many anatomical sites. However, DLAS's black-box nature has limited its widespread clinical implementation. Robust evaluations are required prior to the clinical implementation. In this thesis, we present our comprehensive validation approach for assessing the clinical acceptability of DLAS contours in the male pelvis region for automated prostate treatment planning. We then evaluated the DLAS model's capacity for continuous improvement and generalizability and successfully adopted it in a multi-user environment. Additionally, we provided an implementation workflow for this software that can be used by other clinical users."
Manual reviewing contour is a time-consuming process that is prone to errors and omissions, leading to dosimetric uncertainties and lower quality of radiation treatment. To assist with the manual contour review process, an automated contouring QA tool is necessary. We proposed a machine learning-based methodology for an automated contour quality assurance system that detects errors in manual contouring, using the precise DLAS contour as a reference. Moreover, we established a knowledge-based contour QA system that can localize and categorize contour errors for improved accuracy and efficiency.
Overall, this dissertation provides a more comprehensive understanding of DLAS in a clinical multi-user environment, which will improve the quality and safety of the radiotherapy workflow.
Recommended Citation
Duan, Jingwei, "IMPROVING RADIOTHERAPY WORKFLOW: EVALUATION AND IMPLEMENTATION OF DEEP LEARNING AUTO-SEGMENTATION IN A MULTI-USER ENVIRONMENT, AND DEVELOPMENT OF AUTOMATIC CONTOUR QUALITY ASSURANCE SYSTEM" (2023). Theses and Dissertations--Radiation Medicine. 5.
https://uknowledge.uky.edu/radmed_etds/5
Digital Object Identifier (DOI)
https://doi.org/10.13023/etd.2023.107
