DSA Vascular Enhancement And Center-line Research

Along with the increasingly advanced social economy and living standard, cardiovascular and cerebrovascular diseases are becoming a major threat to human health. The most direct approach to examine whether heart and cerebral vessels are diseased and blocked is to conduct angiography. Digital Subtraction Angiography technology(hereinafter referred to as: DSA) which highlights vascular images and leaves out those of the overlapped bone and soft-tissue system can achieve better image acuity. Therefore, DSA has taken the place of traditional angiography and been widely applied to clinical diagnosis. However, sound judgment is difficult to make only by observing angiography images. The doctors should be provided a lot more vascular information acquired straightly from post-processing DSA images before they adopt more accurate therapeutic plans and minimize medical errors to the fullest extent.Since various methods are used to process DSA images, this paper aims at studying how to enhance them and extract centerlines. Vessel three-dimensional reconstruction is based on its two-dimensional extraction, which enables actual reflection of vascular distribution and structure. Thus, it is of great significance to research into improving extraction algorithms of vascular centerlines.To begin with, images needs preprocessing. In order to enhance the contrast of vascular images to extract their problems, this paper proposes a joint method of segmental linear transformation and multi-scale filter. At first, adjust image greyscale to a reasonable range by means of segmental linear transformation; then build multi-scale filter out of Hessian Matrix according to eigenvalues of Hessian Matrix and multi-scale theory so as to intensify image acuity of vascular structure of different sizes and suppress noise interference. It is shown in the simulation experiment that with the help of the integrated filter this method are supposed to enhance vascular images with greater effect.After that a further study about the extraction algorithms of vascular centerlines is carried out. Firstly, the enhanced vascular images are segmented by thresholds of Ostu algorithm; then segmented connecting regions can remove interference from the noise and porphyritic structure. Secondly, vessel framework, i.e. the coarse centerline, is extracted by open-close operation of morphology. Finally, directions of vessel transversal lines are located by referring to the Hessian Matrix of each point in the coarse lines; and grey extreme value can also be located by using grey projection-fit algorithm on each transversal point. Accordingly, sub-pixel precision of vascular centerlines is exactly obtained. It is shown in this simulation experiment that extraction algorithms of vascular centerlines analyzed in this chapter are not only faster in speed but also more acute in precision, which is definitely important in its application value.