| Coronary heart disease(CHD)is a kind of heart disease which is harmful to human health.Marking the stenosis area of coronary artery accurately in CT images can assist doctors in diagnosis and treatment.In view of this,an automatic detection algorithm of coronary artery stenosis area is proposed,which includes two parts: centerline extraction and stenosis detection.Firstly,in the centerline extraction part,the thinning algorithm based on template deletion and the thinning algorithm based on P-simple points are studied,and an algorithm combining thinning and distance field is proposed.This algorithm innovatively combines distance field algorithm on the basis of template deletion algorithm,and resolves the problem of skeleton disconnection of template deletion algorithm.The main steps include: using template deletion thinning algorithm to get the initial skeleton,finding the skeleton disconnection area,calculating the distance field of the disconnection area,and connecting the disconnection area.After the skeleton line is obtained,it also needs to be converted into the center line required for the subsequent steps.The key point of the algorithm is how to find all the skeleton disconnected areas effectively.Therefore,a variety of detection conditions and subsequent correction methods are designed.Compared with the thinning algorithm of template deletion,the thinning algorithm combined with distance field can effectively connect the broken area of skeleton and maintain the connectivity of skeleton.Compared with the thinning algorithm of P-simple points,the thinning algorithm combined with distance field can extract similar skeleton.However,the combination of thinning and distance field has great time advantage.The experimental data show that the average time of thinning algorithm combined with distance field is 7.458 seconds,and the average time of P-simple point algorithm is 38.395 seconds.Secondly,in the stenosis detection part,a stenosis detection algorithm is designed by using the stenosis measurement formula of the vessel cross-section boundary.The main steps are to segment the vascular tree,calculate the stenosis coefficient of the central point and determine the initial stenosis area,remove the stenosis area of the smal er vessels,expand the stenosis area,and subsequently remove the shorter stenosis area and the stenosis area with closer connection.The key point of the algorithm is how to find the reference radius accurately when calculating the stenosis coefficient and how to optimize the stenosis areas.The experimental results show that the stenosis detection algorithm has a good effect.The correct detection rate of the data given by the cooperative company is 80.65%,and all the more serious narrow areas are detected. |
