Automated Assessment of Pulmonary Arterial Morphology in Multi-row Detector CT Imaging Using Correspondence with Anatomic Airway Branches

Abstract

Multi-row detector CT (MDCT) provides high resolution structural and functional imaging that has been helpful in studying altered physiology, making early diagnosis, and evaluating treatments in pulmonary research. There is growing evidence suggesting that pulmonary vascular dysfunction plays a major role in progression of centrilobular emphysema, a component of chronic obstructive disease (COPD). Few studies have attempted to quantify central pulmonary vessel morphology and to compare these measurements across COPD groups. However, the scope of vascular structures examined in such studies has been limited, primarily, due to lack of an automated and standardized method of comparing matching vessel branches. In this paper, we present a fully automated method, using a novel arc skeletonization and a local correspondence analysis, to identify matching pulmonary arteries by linking those with anatomically defined specific airway branches. This method provides a standardized way of establishing correspondence between matched pulmonary arteries for intra- and inter-subject scans. The accuracy and repeatability of the method was examined on non-contrast MDCT scans of 10 normal subjects. It was observed that 83% of the arteries classified by our automated method agree with “true” arteries as labelled by an interactive manual artery-vein separation tool. Repeat scan intra-class correlation of arterial morphological measures over six anatomic airway branches was observed as 91%.