Degree
Master of Science
Program
Medical Biophysics
Supervisor
Parraga, Grace
Abstract
Chronic obstructive pulmonary disease (COPD) leads to persistent airflow limitation, causing a large burden to patients and the health care system. Thoracic CT provides an opportunity to observe the structural pathophysiology of COPD, whereas hyperpolarized gas MRI provides images of the consequential ventilation heterogeneity. However, hyperpolarized gas MRI is currently limited to research centres, due to the high cost of gas and polarization equipment. Therefore, I developed a pipeline using texture analysis and machine learning methods to create predicted ventilation maps based on non-contrast enhanced, single-volume thoracic CT. In a COPD cohort, predicted ventilation maps were qualitatively and quantitatively related to ground-truth MRI ventilation, and both maps were related to important patient lung function and quality-of-life measures. This study is the first to demonstrate the feasibility of predicting hyperpolarized MRI-based ventilation from single-volume, breath-hold thoracic CT, which has potential to translate pulmonary ventilation information to widely available thoracic CT imaging.
Recommended Citation
Westcott, Andrew, "Texture Analysis and Machine Learning to Predict Pulmonary Ventilation from Thoracic Computed Tomography" (2019). Electronic Thesis and Dissertation Repository. 6080.
https://ir.lib.uwo.ca/etd/6080