Pulmonary Arteries Segmentation and Feature Extraction through Slice Marching

We propose a novel method, slice marching1, for segmenting opacified vessels tree in 3D images (volume data), from multislice computed tomography (MSCT) scans. The method uses fast marching with freezing of boundaries to advance inside the vessel, slice per slice. Large scale features, such as vessel section and curvature, are evaluated for each slice. These features can then be used to influence the speed of propagation (in the sense of interface evolution theory), thus combining image data and anatomical properties for the segmentation task. As a by-product of the method, the extracted features can also be used for other purposes, for instance, vessels diagnosis. Keywords— segmentation; pulmonary arteries; pulmonary embolism; fast marching; region growing; CT; vessels; emboli detection