Author ORCID Identifier
0000-0001-9935-2273
Date of Graduation
5-2024
Document Type
Dissertation (PhD)
Program Affiliation
Medical Physics
Degree Name
Doctor of Philosophy (PhD)
Advisor/Committee Chair
Kristy Brock
Committee Member
David Fuentes
Committee Member
C. Dave Fuller
Committee Member
Stephen Lai
Committee Member
James Long
Committee Member
R. Jason Stafford
Abstract
Head and neck cancer (HNC) is the sixth most diagnosed cancer worldwide with over 600,000 patients diagnosed every year. One of the primary methods to treat HNC is through radiation therapy (RT). While RT is effective at treating HNC, late toxicities resulting from treatment can occur months to years post-treatment and can cause debilitating quality of life changes for HNC survivors. One of these late toxicities is osteoradionecrosis (ORN), which is the death of bone due to radiation. For HNC, the mandible is commonly affected and can cause challenges chewing, swallowing, and physical appearance changes. Several factors indicate the pressing need to reduce ORN such as improving HNC survival rates and a lower age of HNC diagnosis due to the increase of HPV-associated cases. One potential method to achieve this goal is by evaluating imaging biomarkers associated with ORN development. With appropriate imaging biomarkers, HNC treatment can be optimized to reduce the prevalence of ORN and allow for earlier ORN detection.
In this dissertation, several imaging biomarkers were evaluated for their potential use in monitoring ORN development. The imaging biomarkers studied were radiation treatment dose, post-treatment dynamic contrast enhanced magnetic resonance imaging (DCE-MRI), and the change in pre- and post-RT DCE-MRI. In the first project, several machine learning and deep learning methods were compared to predict binary ORN status based on treatment dose. This project found that the machine learning algorithms studied performed similarly to each other and outperformed the deep learning algorithms used. The final model test set area under the receiver operating characteristic curve was 0.70. The next project analyzed high radiation therapy dose (> 60 Gy) in the mandible in post-treatment DCE-MRI. The Wilcoxon signed-rank test in this study determined that there was a statistically significance difference in the DCE-MRI quantitative parameter ve between high dose (> 60 Gy) and low dose (≤ 60 Gy) regions of the mandible (W=214, Z=3.85 p=0.00013, n=48). Finally, in the last study, a pipeline was built as part of an ongoing clinical trial to determine the association between the changes in pre- and post-RT DCE-MRI quantitative parameters and ORN development. This pipeline will be used as part of an ongoing clinical trial to determine if there is a statistically significant difference in the mandibular Ktrans and ve between ORN negative and ORN positive subjects.
This group of studies shows the possibility in using different imaging biomarkers for the early identification of ORN and as a potential tool to aid in ORN intervention. The first project's work can serve as a guide for future studies on the use of machine learning and deep learning for late toxicity prediction. From the second project, the association between post-treatment DCE-MRI and high delivered radiation dose could be used to motivate further studies to understand the relationship between DCE-MRI and dose during treatment. Finally, the results from the final project could inspire future work that analyzes the changes in DCE-MRI during radiation treatment and methods to adapt treatment plans to minimize ORN development.
Keywords
Head and Neck Cancer, Osteoradionecrosis, Radiation Therapy, Dynamic Contrast-Enhanced Magnetic Resonance Imaging, Imaging Biomarker, Deep Learning