Segmentation of Pulmonary Artery Based on CT Angiography Image

This paper presents an efficient algorithm for segmenting pulmonary artery tree in 3D images from Computed Tomography Angiography (CTA) scans. A new algorithm based on region growing and slice marching (RGSM) is proposed to segment pulmonary artery and track the artery tree section. To avoid the leakage and other false positive such as pulmonary vein, thresholds are automatically determined based on histogram of mediastinum, and a feature map based on eigenvalues of Hessian matrix which make the edge between different organs more obvious is constructed to stop the propagation from blowing up the boundary of desired pulmonary artery close to other organs in the whole process. Segmentation of pulmonary artery and labeling of artery tree structure is finally implemented by RGSM algorithm along with morphological operation within the mask after threshold and feature map filter. The proposed method is evaluated on 18 continuous clinical patient data from CTA scans. With regard to acceptability, two qualified radiologists conducted a visual judgment on the segmentation results in search of any gross miss segmentation.