ECCV'06 Proceedings of the 9th European conference on Computer Vision - Volume Part III
Hi-index | 0.00 |
This paper presents a novel approach for segmenting and tracking pulmonary artery tree from mediastinum to lung in computed tomography angiography images. An algorithm based on region growing and slice marching methods is proposed for vessel tree tracking, and the approach also combines with context-contour-based mediastinum segmentation, feature map filter, and 3D morphological techniques to assist the track of pulmonary artery in avoiding leakage problem. We have evaluated our method on 18 continuous clinical patient data and demonstrated its ability to correctly separate pulmonary arteries from other blood vessels, and robustly handle noise.