Automatic Quantification of the Aorta and Pulmonary Artery in Chest CT: methods and validation in lung screening

Cardiovascular diseases and Chronic Obstructive Pulmonary Disease
(COPD) are among the major leading causes of death globally. In the
search for early identification of individuals at risk of cardiovascular
disease in COPD, imaging-based assessments of the shape and size of the
aorta and pulmonary artery have rapidly gained interest. Changes in
these two large arteries may indicate cardiovascular diseases such as
pulmonary hypertension and aortic aneurysm. Furthermore, the ratio of
the diameter of the pulmonary artery to ascending aorta at the level of
pulmonary artery bifurcation is shown to be associated with an increased
risk of severe exacerbations and increased mortality in patients with
COPD. Therefore, it is essential to accurately delineate and quantify the
anatomy of the aorta and pulmonary artery. With the growing use of
low-dose non-contrast thoracic CT scans for lung cancer screening, there
is an opportunity to measure the aorta and pulmonary artery in these
scans. However, performing diameter measurements manually is laborintensive; therefore, automatic 3D segmentation and measurement
techniques are desirable.
This thesis develops and validates fully automatic segmentation and
diameter measurement techniques to quantify the shape and size of
aorta and pulmonary arteries in CT scans. It presents a method based on
optimal surface graph cuts to segment the aorta and pulmonary arteries
separately and extract landmarks for each vessel for automatic diameter
measurement. It also presents a new deep-learning-based approuch
named Posterior-CRF, for jointly segmenting the vessels. The methods
present robust and reproducible results of sufficient accuracy and
reliability for use in the clinical study.