Segmentation of CT angiographic images of the coronary artery based on an improved active narrow band model

Menglu ZHANG,Xiaoping YANG

DOI: https://doi.org/10.3969/j.issn.1005-202X.2017.02.012

2017-01-01

Abstract:As an important step in vascular disease study,segmentation of the coronary artery images is closely associated with stenosis grading and plaque detection.Herein we propose a new method for segmentation of CT angiographic (CTA) coronary artery images by combining the active narrow band and sign pressure function (SPF) level set to solve the problems of gray scale heterogeneity and low contrast of the vessels.Morphological dilation and erosion operations were applied to the initial contour to establish an active narrow band to limit the contour evolution.A local SPF was then constructed in the active narrow band,and a level set algorithm was used to facilitate the convergence of the initial contour into the exact contour.Finally,a morphological closing operation was utilized to smooth the curve.With the use of active narrow band to localize the image region,the computational complexity and the gray scale heterogeneity of the images were reduced to promote the evolution of the contour curve to the tiny peripheral vessels and narrow areas.The experimental results showed that compared with the traditional segmentation method,the proposed method achieved more accurate segmentation of the coronary artery images to facilitate the diagnosis of the vascular lesions.