The Segmentation Methods Of Coronary Angiogram Images

The heart cerebrovascular disease is the main threats to the human health currently. Now there are lots of diagnostic approaches about cardiovascular disease, angiography is the most reliable method of all. For angiography, the contrast medium is injected to the blood stream, then the heart and great vessels are developed to X-rays so that contraction and relaxation function of the heart can be clearly observed. The doctor can observe whether the coronary arteries are narrow,whether the blood clot exists by angiography images. Then the doctor can decide whether the blood circulatory system and the heart are normal. So, the angiography images can provide more and more accuracy information than other diagnostic methods. The coronary angiogram segmentation is an important application about image segmentation technologies in the medical field. After the image being segmented, the structure of vessels can be widely applied, such as assisting the doctor to diagnosis, quantitative analysis on vascular tissue, accurate positioning the narrow part of vessels and directing the doctor to operate and so on. Considering the importance of coronary vessels segmentation, in the paper the feature of coronary angiogram images is analyzed and the segmentation methods are discussed. The main works and innovations include:1. The research on the segmentation methods of coronary angiogram images based on fusion.The two popular methods about coronary angiogram segmentation are coronary arteries segmentation based on morphologic method and coronary arteries segmentation based on Gaussian filter. The morphologic top-hat method can partly enhance the grey of vessels. However, the grey enhancement of tiny blood vessels is not obvious due to the similarity between the grey of tiny blood vessels and the grey of background, so that the tiny blood vessels will be removed as background. The grey feature of vessels is considered in the segmentation methods based on Gaussian filter which can extract the major blood vessels and tiny blood vessels efficiently. However, it also extracts noises which are similar to the size of tiny blood vessels. In the paper, the results of two segmentation methods are analyzed by the region connectivity and the segmentation method of coronary angiogram images based on fusion is proposed. At first, the top-hat method and Gaussian filter method are used to enhance the same coronary angiogram, then two enhanced images can be obtained. After that, the method of optimal entropy is used to segment blood vessels and two images containing coronary arteries can be derived. At last, two images are fused and blood vessels are extracted. Results show the method can extract major blood vessels and tiny blood vessels efficiently, and the noises can be removed at the same time.2. The research on the segmentation methods of coronary angiogram images based on local entropy information measure.The extraction and segmentation methods based on transition region are a threshold segmentation reported currently. The traditional exaction method of transition region is an indirect extraction method by clip transformation of grayscale and the average of gradient. Gradient is the feature parameter for extracting transition region, which is so sensitive to noises that transition region sometimes can not be extracted. In the paper, the characteristics of coronary angiogram transition region is analyzed. The key to the transition region extraction is the selection of the feature parameters. According to the characteristic of transition region, the local entropy information measure is constructed as feature parameters for extracting transition region and the two segmentation methods of coronary angiogram images based on local entropy information measure are proposed. The first approach based on threshold segmentation of transition region extraction uses local entropy information measure to extract the transition region of angiogram and determines segmentation threshold by histogram of transition region. This method can extract the trunk of blood vessels properly without good result for extracting the tiny blood vessels. So, the fusion method based on transition region is proposed. For this method, the major blood vessels and the transition region of images are extracted firstly, then using the difference of blood vessels and noise fragments, blood vessels are extracted by analyzing the region connectivity. The major blood vessels and tiny blood vessels can be extracted efficiently. The connectivity among tiny blood vessels is better than other methods mentioned in the paper..3. The research on the segmentation methods of coronary angiogram images based on local complexity information measure.Local entropy is an important parameter of local entropy information measure which can denote the grayscale changing frequency of image local neighborhood. Computing local entropy is a complicated and difficult process, local complexity is a statistics about the grey level change in image local neighborhood. Local entropy can be taken place by local complexity as the parameter of image grey change. In the paper, the characteristics of local complexity is analyzed and the viewpoints of informatics are combined to construct the new feature parameters for extracting transition region. Two segmentation methods of coronary arteries are proposed by using local complexity information measure. Threshold method can extract the trunk of blood vessels without good extraction effection for tiny blood vessels. The fusion method based on transition region extraction can improve the extraction of tiny blood vessels and remove the background noises. Comparing the method based on local entropy information measure to the method based on local complexity information measure, the results show the methods based on local complexity information measure can reduce computational complexity and save computational time.4. The research on the segmentation methods of coronary angiogram images based on neighborhood unhomogeneity of degree information. The method of image segmentation based on graph theory is a new research hotspot in the field of image segmentation, currently. Graph theory is a branch of mathematics, its research object is graph and the graph is regard as a figure constructed by some points and some lines which connect two points. An image can be mapped into a weighted undirected graph, the pixels in the image can be looked as the nodes of graph and the edges can be formed between every pair of nodes. The degree of nodes is an important parameter of undirected graph which is decided by the weight of edges linking to nodes. Considering of the poor connectivity of coronary arteries extracted by existing segmentation algorithm, graph theory is introduced to angiogram segmentation. According to the characteristics of degree information, another measurement parameter of image transition region extraction is constructed which is the neighborhood unhomogeneity of degree information. In the paper, two segmentation methods of coronary angiogram images based on degree information neighborhood unhomogeneity are proposed after exacting the transition region. The two methods improve the quality of tiny vessels extraction and the connectivity of tiny blood vessels on different degree.