Improved GAC Model-based Pulmonary Artery Segmentation of CTPA Image Sequence

Abstract

Aiming at the problem that the traditional geodesic active contour (GAC) model is prone to produce boundary leakage and cannot segment adaptively, this paper proposes an improved GAC model and realizes the automatic segmentation of pulmonary artery in computed tomographic pulmonary angiography (CTPA) image sequence. Firstly, the variable velocity C(I) is used to replace the constant velocity c of the traditional GAC model, and base on the improved GAC model with C(I) to segment the first frame image of the CTPA sequence to obtain a convergent pulmonary artery contour. Secondly, the grayscale information of the target area is used to improve C(I) to V(I), which makes its direction variable. Finally, based on the improved GAC model with V(I), automatic segmentation of pulmonary artery in subsequent images is realized. Among them, the initial contour of each subsequent image is the final contour of the previous image. These two improvement strategies can solve the problem that the model is easy to be over-segmented and drive the curve evolve adaptively inward or outward to the target contour. Experimental results show that the proposed algorithm can realize automatic segmentation of pulmonary artery, and has a high coincidence rate with the results of physician segmentation.