Research On Non-rigid Registration Method For Cerebrovascular Digital Subtraction Angiography

Digital Subtraction Angiography (DSA) has got broad application in clinical diagnosis and treatment. It is an irreplaceable and effective vas visualization tool. And cerebrovascular DSA is a necessary part in interventional operation for cerebrovascular pathological changes. How to improve the imaging function and quality in providing powerful gist for clinical diagnosis and treatment is one of the hotspot problems in medicinal image processing domain. The registration matter in cerebrovascular DSA is systematically researched in this paper. An effective non-rigid registration algorithm for cerebrovascular DSA is presented. On the peculiar flows in non-rigid registration, the following study works are carried out:Firstly, the algorithm based on regular girding model is adopted to build serial control points. The temporal efficiency of this approach is superior to the method based on image character evidently. The automation level of clinical application for cerebrovascular DSA registration is advanced observably. The algorithm based on Energy of Histogram of Differences is adopted to accomplish the comparability mensuration of matching operation. The experiment result bears out that this rule is the same with the registration between Mask Image and Live Image greatly. Whether there is vas shadow in the rectangular matching template after angiography or not, the active displacement can be found exactly and fleetly.Secondly, for the trait of the control points' form, the Powell algorithm based on fixed template is utilized to optimize the searching for correlative points. The searching range of control point in planar space is restricted so as to avoid plunging into local optimal solution.Afterwards, according to the characteristic of cerebrovascular DSA , an algorithm based on positive stretching spatial transformation is introduced to solve the distortion of matching template, and to achieve the activity estimation of control point in matching process. An algorithm of pixel mapping based on inverse stretching spatial transformation is produced to create the target image of Mask Image after distortion, and to gain an exact subtraction image.Finally, on the base of DSA image registration strategy in this paper, a special software for cerebrovascular DSA image non-rigid registration is exploited. It lays a solid foundation for the coming practical clinical application.