Electronic Thesis and Dissertation Repository

Degree

Doctor of Philosophy

Program

Medical Biophysics

Supervisor

Dr. Aaron Ward

2nd Supervisor

Dr. David Palma

Joint Supervisor

Abstract

Stereotactic ablative radiotherapy (SABR) is a guideline-specified treatment option for patients with early stage non-small cell lung cancer. After treatment, patients are followed up regularly with computed tomography (CT) imaging to determine treatment response. However, benign radiographic changes to the lung known as radiation-induced lung injury (RILI) frequently occur. Due to the large doses delivered with SABR, these changes can mimic the appearance of a recurring tumour and confound response assessment. The objective of this work was to evaluate the accuracy of radiomics, for prediction of eventual local recurrence based on CT images acquired within 6 months of treatment. A semi-automatic decision support system was developed to segment and sample regions of common post-SABR changes, extract radiomic features and classify images as local recurrence or benign injury. Physician ability to detect timely local recurrence was also measured on CT imaging, and compared with that of the radiomics tool. Within 6 months post-SABR, physicians assessed the majority of images as no recurrence and had an overall lower accuracy compared to the radiomics system. These results suggest that radiomics can detect early changes associated with local recurrence that are not typically considered by physicians. These appearances detected by radiomics may be early indicators of the promotion and progression to local recurrence. This has the potential to lead to a clinically useful computer-aided decision support tool based on routinely acquired CT imaging, which could lead to earlier salvage opportunities for patients with recurrence and fewer invasive investigations of patients with only benign injury.

Download

Share

COinS