Three-dimensional Reconstruction Of Angiography

Cardiovascular disease is mainly caused by insufficient blood flow to the heart muscle heart beat obstacles. Now it has become a threat to human health, the first of the major factors, with a high prevalence, high mortality rate and its causes is coronary atherosclerosis or spasm caused by vascular stenosis or occlusion. Therefore, it is particularly important for the early diagnosis of cardiovascular diseases.Conventional cardiovascular disease diagnosis method is the use of digital subtraction angiography(DSA). It is also the diagnosis of cardiovascular disease as the "gold standard", the doctor can in real time from two-dimensional images of DSA see vessel stenosis. However, conventional DSA image only from two angles of view of disease of observation, diagnosis, a lack of focus on the three-dimensional diagnosis function, which in a certain extent limits the comprehensiveness of DSA in the diagnosis of. In addition, the current diagnosis of vascular imaging currently depends on the doctor's subjective consciousness and clinical lack a on vascular stenosis disease such as automatic analysis of adjuvant diagnostic tool. The research of a method that can accurately segment the coronary artery from the DSA image sequence has important practical significance and clinical value.In DSA images, the distribution of coronary artery is not concentrated and interfered by other organizations, which brings great challenges to the segmentation of the coronary artery. Based upon surveying medical image processing and registration algorithm, according to the coronary artery image features, this paper proposes local feature registration method suitable for DSA scene inspection. Based upon this method, it is able to obtain accurate registration result. In order to realize 3D modeling of DSA data, this paper uses homography assumption to conduct intensification processing on the registration result, and then uses computer vision's “structure from motion” theory to conduct 3D modeling according to the calculated intrinsic parameters and extrinsic parameters in the camera.This paper mainly carried out the following research work:Vascular enhancement. First, the image pretreatment, to suppress the lung region; then using single Hessian matrix enhancement and filtering to improve the image of coronary artery area and the surrounding tissue contrast, for the overall regional growth provides good conditions; finally, use of the ascending aortic position information as the seed point, on the ascending aorta and enhanced image fusion of data overall regional growth. The segmentation result of the coronary artery can be obtained by the difference of the segmentation results of the ascending aorta.Depth calculation. Using theoretical structure of computer vision motion restoration, in the local feature points registration based on using the theory of multi affine transform(multi affine transform) to strengthen the registration results and obtain the ultimate secret set of matching points. Finally, according to the calculation of the camera parameters and external parameters for three-dimensional modeling.Optimization of three dimensional depth. To construct the initial 3D depth information usually has a larger error, which is mainly due to caused by image matching error with punctuality. In this paper, a three-dimensional depth optimization method is designed for the 3D modeling of vascular DSA, so that the final 3D modeling results can meet the clinical needs.