Pixels to Parameters: Automated Bronchial Tree Analysis in Low-Dose CT for Normative Lung Imaging Biomarkers

The airways are tubular structures that bring air to the lungs for delivery of oxygen to the body, and removal of waste gasses like carbon dioxide. In cases of respiratory illness such as asthma or chronic obstructive pulmonary disease (COPD), the airways undergo changes. Measuring these changes could allow us to catch early signs of disease or to monitor treatment of established disease. However, to better understand changes in airways related to disease, we need to first understand how normal airways appear in the general population.
This thesis uses advanced computer analysis of CT scans to measure the airways in a large group of mostly healthy individuals. We developed an automated method combining artificial intelligence and computer vision that separates out the full 3D structure of the airways from low-dose CT scans. The goal of this thesis is to understand how the airway measurements vary in healthy lungs and how factors like age, sex, height, weight, and smoking history affect them. Until now, studies on airway measurements mostly focused on people with long smoking history or with lung diseases like COPD. The results of this thesis provide reference values for airway measurements that other studies can compare to. Results show that age, sex, height, weight, and smoking all affect airway measurements on a CT scan, and so these factors must be considered in future studies.