Pulmonary Artery Segmentation Based on Three-Dimensional Region Growth Approach

Abstract

Segmentation of pulmonary artery (PA) and its branches is the primary work to detect pulmonary embolism (PE). In order to achieve precise segmentation of the pulmonary artery, we propose an improved three-dimensional region growth (3D RG) method based on the anatomical features and spatial connectivity of the pulmonary artery. The three-dimensional vector is extracted from pulmonary trunk according to the mathematical theorem, and the isolation plane is introduced to prevent segmentation leakage. By adding an iteration parameter, the branches are segmented exactly. In this paper, three-dimensional segmentation and reconstruction of the pulmonary artery and its branches are carried out with pulmonary CTPA images provided by the China-Japan Friendship Hospital. The improved algorithm can effectively segment the pulmonary artery and its branches to the lung subsegments. High accuracy and three-dimensional connectivity are guaranteed and the research achieves a desired effect. According to the evaluation of radiologist, the segmentation result is accurate.