Coronary Artery Segmentation Algorithm For CTA Images Based On Pre-extraction Of Heart Area

In recent years,cardiovascular diseases have become the biggest threat to human health,with high morbidity and high mortality.With the improvement of medicial level,many diagnostic imaging technologies for cardiovascular diseases have been developed clinically,which can help doctors to directly analyze and diagnose the disease.Among them,Computed Tomography Angiography has become one of the most commonly used early screening methods for coronary heart disease due to its low price and non-invasive diagnosis.At present,the amount of CTA medical image data is growing rapidly,but the training process of medical imaging experts is slow.It is extremely complicated and difficult for experts to manually complete large-scale CTA image segmentation.If automatic segmentation can be adopted,it will bring them different degrees of help.In this paper,the coronary artery CTA image is taken as the research object,and the blood vessel segmentation is realized from a three-dimensional perspective.The research work is mainly divided into the following four parts:(1)A complete system of phased coronary CTA image segmentation is designed.Firstly,an adaptive threshold method is used to pre-extract the heart region for rough segmentation of blood vessels,which provides a good initial value for deep learning networks.Secondly,the heart region extracted by the traditional algorithm is combined with the corresponding label as the training data of the V-net network to achieve fine coronary artery segmentation.In the final optimization phase,the level set function is used to iterate smooth the blood vessel edge contour to obtain the final blood vessel segmentation result.(2)This paper uses the adaptive threshold algorithm to extract the heart region in the CTA image during the rough segmentation of blood vessels.Because the coronary arteriesin the CTA image are not concentrated and the topological structure is complex,and the image is prone to problems such as uneven grayscale,low contrast,and closeness of grayscale between different tissues,the CTA image segmentation has become a difficult problem in medical image processing.Based on the characteristics of coronary CTA images,this paper proposes an adaptive threshold method to pre-extract the heart region and the coronary vessels are segmented into a relatively clean region at the stage of rough segmentation of blood vessels.At the same time,several classical algorithms of traditional blood vessel segmentation are researched and implemented,and the rough segmentation algorithm of blood vessels is applied to the traditional algorithms,the effectiveness of the pre-extraction method of the heart region is compared and analyzed.(3)In the stage of fine segmentation of blood vessels,this paper selects deep learning network for three-dimensional segmentation of coronary arteries.In order to achieve high-precision automatic segmentation of coronary blood vessels,after studying the core ideas of deep learning for semantic segmentation,this paper proposes a complete segmentation algorithm based on deep learning,including network selection,network improvement,and training test methods.The visual comparison between the segmentation result of the improved V-net Network and the classical Fully Convolutional Network(FCN)is made.(4)This paper uses the level set function to post-process the network segmentation results in the vascular optimization stage.Aiming at the problem of fuzzy contours of blood vessel edges predicted by the network,this paper adds a level set function to iteratively optimize the edge contours of the network predicted results to further improve the accuracy and smoothness of the results of blood vessel segmentation.Compared with the experimental results of FCN,the method proposed in this paper has an average increase of 6.4% and 8% in the coefficient of Jaccard and the coefficient of Dice.The experiment proves that the method proposed in this paper has good effectiveness and practicability,and can accurately segment coronary artery CTA in three dimensions,which is helpful for the diagnosis and treatment of coronary heart disease,and provides doctors with the role of auxiliary diagnosis.