PhD Defence: Improving the Understanding of Structural and Biomechanical Changes in Chronic Obstructive Pulmonary Disease (COPD) Using Expiration CT-derived Texture-Based Radiomics

Student: Meghan Koo

Supervisor: Dr. Miranda Kirby

Abstract

Expiration and biomechanical metrics derived from computed-tomography (CT) imaging, such as gas trapping and the Jacobian determinant, show abnormalities in local lung expansion and contraction in patients with Chronic Obstructive Pulmonary Disease (COPD), and are related to respiratory morbidity and mortality. However, they are global measures and do not capture any regional heterogeneity.

In this dissertation, we have developed and introduced novel CT metrics extracted from expiratory and paired inspiratory-expiratory CT texture-based radiomics in COPD, and have investigated the association between these features and important COPD outcomes, such as lung function and disease progression. We have shown that texture-based radiomics features extracted from expiratory and paired inspiratory-expiratory CT images will be able to improve prediction model performance by providing independent and complementary information to conventional CT measurements.

Specifically, we aim to determine if the addition of expiration CT texture-based radiomics features to existing gas-trapping measurements improves model performance for lung function decline. In addition, we aim to develop an approach to quantify the heterogeneity in local lung mechanics using CT texture-based Jacobian radiomics, and to evaluate the model performance of Jacobian texture-based radiomics with lung function and its decline. Lastly, we aim to evaluate the predictive performance of second-order Jacobian-derived texture-based radiomics for CT emphysema progression over 3-years in comparison with conventional CT Jacobian and emphysema measures.

Overall, texture-based expiratory-CT derived measurements such as the Expiratory RadScore and second-order Jacobian-derived texture-based radiomics features may offer opportunities for early disease detection, subtype classification, prognosis, and prediction of various clinical outcomes, thus complementing traditional diagnostic tools and addressing unmet clinical needs.