Locally adaptive fuzzy pulmonary vessel segmentation in contrast enhanced CT data

Abstract

Pulmonary vascular tree segmentation is the fundamental basis for different applications, such as the detection and visualization of pulmonary emboli (PE). Such an application requires an accurate and reliable segmentation of pulmonary vessels with varying diameters. We present a novel fuzzy approach to pulmonary vessel segmentation in contrast enhanced computed tomography (CT) data that considers a radius estimate of the current vessel to adapt the segmentation parameters. Hence, our method allows to capture even vessels with small diameters while suppressing leakage into surrounding structures in close proximity of vessels with large diameters. The method has been evaluated on different chest CT scans of patients referred for PE and demonstrates promising results. For quantitative validation, randomly selected sub-volumes that have been semi-automatically segmented by a medical expert have been used as reference to compare the locally adaptive method against the same method with global parameters.