Electronic Thesis and Dissertation Repository

Degree

Doctor of Philosophy

Program

Medical Biophysics

Supervisor

Dr. Eugene Wong

Abstract

Despite modern treatment advances, overall survival (OS) remains poor for many cancers such as liver and brain. Cancer is a fundamentally heterogeneous and adaptable disease and therefore personalized adaptive treatment strategies may be a key towards improving OS. Radiotherapy, a commonly used cancer treatment technique which employs ionizing radiation to kill tumours, holds promise for delivering adaptive treatment. However, effective adaptation requires the ability to assess and predict tumour treatment response. Therefore development of treatment response prediction tools represents a critical first step towards improving patient outcomes via treatment adaptation. The overall goal of this thesis is to develop treatment response prediction methods with a view towards guiding adaptive radiotherapy.

First, we investigated the relationship between radiation dose and local tumour control among patients with primary and metastatic colorectal liver tumours. We established and compared their dose-response relationships and found that 84 Gy and 95 Gy of radiation could provide 90% probabilities of 6-month local control for the primary and metastatic groups respectively.

Tumour control most often cannot be improved simply through escalating the dose to the entire tumour due to increased risk of side effects. However, it may be possible to safely increase the dose to tumour sub-volumes. Therefore, the second and third contributions of this thesis involve development of image-based treatment response prediction methods which are needed to identify tumour sub-volumes where additional radiation should be deposited to improve tumour control.

Our second contribution involved augmenting a voxel-based method known as parametric response mapping (PRM) to account for image registration error (IRE). The augmented PRM helped to quantify and visualize IRE-related variability. In our third contribution, we further generalized PRM to permit collective analysis of multi-parametric image data. The proposed method was applied to multi-parametric imaging from a patient cohort with glioblastoma and was found to predict OS ≥ 18 months (median OS) with a sensitivity and specificity of 90% and 78% respectively.

In summary, these contributions provided some of the response assessment groundwork needed to guide adaptive RT. Image-based dose response relationships via the augmented and multi-parametric response maps will facilitate personalization and guidance of adaptive radiotherapy.


Share

COinS