[en] Cardiovascular disease (CVD) is the most common cause of death in the world. Accurate coronary arteries segmentation is the basis for heart disease diagnose. But in coronary computed tomography angiography(CCTA) images, the complex structure of coronary vessels and the low contrast of thin vessels cause difficulty in coronary arteries segmentation. In this paper, an automatic segmentation algorithm is proposed to extract vessels from CCTA. Firstly, the hessian matrix is used to enhance vessels in CCTA images, and then, the moving sphere model we proposed is used to extract centerline of coronary arteries. Experimental results showed that the proposed model provides good-quality segmentation. (paper)