Tracking Based Coronary Artery Centerline Extraction Algorithm In X-ray Coronary Angiography

X-ray coronary angiography has been regarded as the most common way to imaging coronary artery and the gold standard of diagnosing coronary aneurysm and coronary artery stenosis due to its characteristics such as the low price for examining,high resolution for imaging,low technical barriers and so on.The centerline of coronary angiography is one of important prerequisites for many medical image analysis applications and computer aided diagnose systems.On the other hand,centerline of coronary angiography can play an important role in many clinical applications directly.For examples,the centerline of coronary artery can be used in coronary artery segmentation,registration of coronary artery in different perspectives and imagine modalities,heart motion analysis,calculating SYNTAX score and so on.Therefore,it is a meaningful thing to extracting centerline of coronary artery for X-ray coronary angiography.Tracking based centerline extraction algorithms have become one of the mainstream algorithms for extracting centerline in X-ray coronary angiography due to its several advantages such as the easy implementation,short running time,having the ability to extract the complete centerline of whole coronary artery tree easily and so on.However,there are still some shortcomings about the published tracking based centerline extraction algorithms,such as the results of centerline can be affected by the random noise and salt-and-pepper noise easily,bad performance in extracting centerline when the angiography with uneven contrast medium,cannot adjust the centerline point with the boundary information of coronary artery and so on.Therefore,there are many jobs can be finished to improve the performance of tracking based centerline extraction algorithms.In light of the mentioned problems of tracking based centerline extraction algorithms,the research content of this thesis is showing as follow:Firstly,a tracking based centerline extraction algorithm is proposed in this thesis,and there are five septs in our algorithm.(1)Image preprocessing.The X-ray coronary angiography are preprocessed by using noise adaptive fuzzy switching median filter(NAFSMF)and contrast limited adaptive histogram equalization(CLAHE)to filter the random noise and adjust the image contrast.(2)Selecting seed points for tracking.Selecting rigid points from angiography by using Ridge Point Existence Theorem at first,and then the selected rigid points are fused with the angiography which was enhanced by Frangi filter to filter out the useless rigid points that located out the range of coronary artery.Finally,the filtered rigid points are sorted by the response to Frangi filter and choice the top 20 points as the seed point.(3)Finding out the initial tracking direction.Finding out the initial tracking direction based on the local gray value.(4)Adaptive tracking.According to the tracking direction,the approximate position of the coronary artery centerline is determined by searching within a certain range of arc length with a certain step.The coronary artery centerline is corrected by combining the edge information from the narrow band level set,and the direction and searching step of the blood vessel tracking are updated.(5)Stopping tracking.Stop tracking when conditions are met.Secondly,in light of the limited researching about dividing coronary artery tree to calculate SYNTAX score,an algorithm about dividing coronary artery tree into several segment based on the extracted centerline is proposed in the thesis.The dividing algorithm is achieved by using the ending points,bifurcation points and topological analysis method.Finally,in order to make this algorithm available for more people,this thesis also developed a coronary angiographic centerline extraction software based on C# and Open CV.The simulated and clinical angiography images are used to test the performance of our algorithms.By introducing some general error measurements,the distance error and area error of this algorithm are calculated quantitatively using simulated coronary angiographic images,which are 0.318 and 0.322 respectively.Through comparison with other similar algorithms,it is found that the proposed algorithm achieves the same accuracy as the current advanced algorithms.The experimental results of real clinical images show that the proposed algorithm is insensitive to random noise and uneven distribution of contrast medium.Based on the extracted blood vessel centerline,the whole coronary artery is segmented from coronary angiography by using the method of region growing.At the same time,the segmentation of coronary tree is realized by using the endpoints and bifurcation points of coronary tree.Finally,the image processing software designed in this paper has a friendly interface,and users can easily extract the centerline of coronary artery.In a word,the algorithm proposed in this thesis is insensitive to random noise and uneven distribution of contrast media.It can use the boundary information of blood vessel to correct centerline points,so that the accuracy of extracting the centerline of coronary artery is higher.Based on the blood vessel centerline extracted in this paper,the proposed segmentation algorithm of coronary tree can realize the segmentation of coronary tree.