| Summary: One of the most important factors of human deaths in the world is heart disease. X-ray angiography is used as a reliable method in the examination and treatment of coronary vessels. The vessels must be correctly segmented to ensure that the vessel is being examined for occlusion and correctly diagnosed by a physician. However, the X-ray used in imaging and the opaque material injected into the blood is used in low rates, because of the fact that they are harmful to the human health. For this reason, the image quality obtained by angiogram is low. In addition, there are undesired organelles and non-uniform brightness distributions on the image that complicate the segmentation procedure. In this study, a novel method has been proposed to remove the factors that cause difficulties on segmentation. By the proposed method, all the frames of the angiogram image sequence were used to determine the vessel regions. To determine the vessel locations, the brightness variations between image frames are calculated. With the proposed “Largest Differences” method, the regions that have maximal changes are determined. Frangi, Matched and Gabor Multi-Scale filters were used to enhance the visibility of vessel structures in the images. Threshing has been done for the segmentation process. Adaptive thresholding methods OTSU and P-Tile are used as the thresholding methods. In the segmentation of the whole image, Gabor filtering and P-Tile thresholding obtained the average accuracy of 92.49%. In the vessel segmentation achieved by the proposed method, 93.81% accuracy value was obtained by Frangi filter and OTSU thresholding.
Keywords: X-Ray Angiography, Vessel Segmentation, Multi-Scale Filtering, Foreground Detection from Image Sequence |
