A quality control tool for HDR prostate brachytherapy based on patient-specific geometry and stochastic frontier analysis
|Authors:||Moosavi Askari, Reza|
|Advisor:||Beaulieu, Luc; Archambault, Louis; Poulin, Éric|
|Abstract:||This thesis focuses on developing an experience-based quality control (QC) tool, derived from the concept of stochastic frontier (SF) analysis in economics and based on patient-specific geometric knowledge to improve the quality of the high-dose-rate brachytherapy (HDR-BT) treatment for prostate cancer. One hundred clinical HDR prostate BT plans, using the transrectal ultrasound (TRUS) asthe only imaging modality, all treated with a single fraction of 15 Gy, and made using Inverse PlanningSimulated Annealing (IPSA) algorithm, are studied. Also, Oncentra Prostate system is employed as the real-time 3D prostate imaging and treatment planning system. American Brachytherapy Society dose parameter recommendations for target and organs at risk (OARs) were followed. Relationships between all the different geometric parameters (GPs) and the four dosimetric parameters (DPs) V100 of the prostate, V75 of the bladder and rectum, and D10 of the urethra were examined. Geometricinformation of the patients and catheters are considered as different GPs. To find the dominant GPs in a given SF model, monotonic relationships between the GPs and DPs of interest are measured using a nonparametric approach: the Spearman correlation coefficient. Then, to determine the optimal SF model for each of the target production SF, and the OARs cost SF models, brute-force searches are performed. Different SF models including all the possible combinations of the dominant GPs in the SF model under study are optimized. Optimization is done using a maximum likelihood method implemented in the statistical computing package R, along with its Generalized Simulated Annealing algorithm. The likelihood ratio test and its corresponding p-value are used to compare the statistical significance of adding new GPs to SF models. A production SF (PSF) model for the target, and a costSF (CSF) model for each of the bladder, rectum, and urethra are developed for TRUS-guided HDR-BTtreatment. The difference between the dose value of a plan obtained by IPSA and the one predicted by an SF model is explored. Additionally, to verify if each of the models developed for the TRUS-guided category of the HDR-BT treatment for prostate is universal, we apply it on another category of HDR-BT treatment, in which computed tomography (CT) was used as the imaging modality. So, a different cohort of one hundred clinical CT-guided HDR-BT plans is taken into consideration. A target production SF and three OARs cost SF models are developed for the CT-based plans. Subsequently, the built-in SF models for the TRUS-based and CT-based plans are compared.|
|Document Type:||Mémoire de maîtrise|
|Open Access Date:||28 October 2019|
|Collection:||Thèses et mémoires|
All documents in CorpusUL are protected by Copyright Act of Canada.