Two-Dimensional Information Extraction Of DSA Images Of Cerebrovascular Vessels

The human health and quality of life have been seriously damaged by cerebrovascular disease. Angiography is the most common and effective vascular spection method. However, due to the influence of the tissues and organs around the vessel, conventional angiography is not conductive to the observation of doctors, which will effect the accuracy of diagnosis. Digital Subtraction Angiography (DSA) can eliminate the impact of the background such as bones and provide with clear information of blood vessels and plays an important part in the present techniques of clinical diagosis and interventional treatments of vascular disease. Hardware technology of current DSA system has been developed to its acme, the promotion of image post-procession ability therefore is the new direction of DSA system development.The methods of DSA image processing have been classified into two dimensional information extraction of blood vessels and three dimensional reconstruction of blood vessels, in which the latter method is based on the former one. Therefore the method of two dimensional information extraction of blood vessels is of great significance.In this study, we chose the most typical cerebrovascular angiography as our research object, the principle of DSA imaging was firstly analyzed, and then a logarithmic subtraction algorithm of variable coefficient logarithmic subtraction algorithm was proposed to achieve the abstraction being carried on the images of pre- and post- angiography images(mask images and live images) and to remove most of the influences of the non-vascular structures.Consequently, in order to remove the noises that produce the fuzzy gray in the cerebrovascular substraction image, we studyed the segmentation method based on the fuzzy morphology. The substration image fuzzy set and the complementary set were then respectively filtered by the opening operation in the fuzzy morphology to achieve the smooth between vessels and their own background. According to the characteristics of consistency of the vascular directions, one final filtering results was selected and given a threshold value from the two filtering results mentioned above.After that, the extraction of vascular centers was studied from the aspects of binary images thinning and image filtering based on Gabor respectively and an improved binary image thinning algorithm and axial scanning methods were then proposed.Finally, identifications of the characteristic points was performed for the extracted centerline; the error points that identified by the conventional method were effectively removed through 4 added templates of judgement.