Research And Implementation Of Vessel Segmentation Algorithm Based On Hessian Matrix And Region Grow

Currently vascular disease has become one of the most important disease influence human health and threat to human life. With the development of medical devices and imaging technology, medical image such as magnetic resonance angiography, X-ray angiography, CT angiography and digital subtraction angiography provide a lot of pathological information, and become an indispensable means of treatment for vascular disease, are widely used in clinical diagnostic procedures. The visualization of three-dimensional blood vessels can not only understand the physiology, anatomy structure and blood flow, but also plays an important role in the diagnosis of vascular disease, surgical planning and guiding therapy. In recent years, vessel segmentation technology has become a key research topic that domestic and foreign research scholars focus on.This thesis mainly research in vessel enhancement and segmentation algorithm of the medical images. In terms of vessel enhancement, use the prior knowledge of the blood vessel image imaging, proposes a new vessel filter enhancement algorithm based on Hessian Matrix. Firstly construct the Hessian matrix of image and calculate the Hessian matrix eigenvalues. Then analyze the eigenvalues and combine with the bi-Gaussian kernel function to construct multiscale linear filter. Finally use the filter to enhance the image. In terms of vessel enhancement, this thesis proposes a region growing vessel segmentation algorithms based on spectrum information. The algorithm firstly makes the Fourier transform to a region of interest containing the vascular structure, obtain the spectral of it. Then analyze the image spectrum, via the information of spectrum to extract the main characteristic direction of local vascular structures. And then combine the edge information to identify the center of vascular structures, and treat it as a seed point of the region growing algorithm. Finally use a modified region growing segmentation algorithm to extract the entire vessel.Through the experiments of different regions including liver blood vessels, abdominal aorta and retinal vascular can show that the proposed methods are effective. Experimental results show that the proposed vessel filter enhancement algorithm not only can protrude the vascular structures effectively, but also suppress the interference of noise and other structures perfectly, is a high-quality, high robustness vessel enhancement algorithms. And the proposed vessel segmentation algorithms, extracts the target area of the vessel efficiency with high quality, but also effectively reduces the manual intervention, has good robustness.