Research On Extraction Of The Coronary Arterial Tree In Coronary Angiograms

Coronary atherosclerotic heart disease is a heart disease caused by vascular stenosis or occlusion because of coronary atherosclerosis, or coronary functional changes leading to hypoxia-ischemia or necrosis of myocardial blood. Coronary heart disease is a global problem of human health, which has become a disastrous disease. Angiography is the process where the contrast medium is injected to the blood stream, and under the X-ray the vessel image can be developed, so that change of the vascular geometric shape can be clearly observed. In recent years, coronary angiography is widely used in clinical diagnosis of coronary heart disease, which is regarded as the "gold standard" of the diagnosis of coronary heart disease. After the image being segmented, the structure of vessels can be widely used, such as assisting the doctor to diagnosis, evaluating severity of coronary artery, and it is the basic prerequisite reconstruction of three-dimensional vascular structure. Based on the coronary angiograms, the feature of coronary angiogram images is analyzed and the performance of extraction of the coronary arterial tree is improved. In this paper, the main works are as followed:(1) A combination of top-hat and Gabor filtering method was proposed to extract the coronary artery region. Based on the imaging characteristics of the coronary angiography and the gray features of blood vessels model, we improved a combination method of morphological methods and multi-scale Gabor function filtering to segment the coronary angiography image. Firstly, we enhanced a coronary angiogram by using morphological top-hat transformation, and segmented the image by using the adaptive threshold. Secondly, we enhanced the coronary angiography image by using the Gabor function filtering, and segmented the image by using the adaptive threshold. Finally, according to the two segmented images, the vascular connectivity between the two images was determined, and attached the vessels by using the Gabor filtering method to the main vessels by using the top-hat transformation, then the segmented image was produced. The experiments showed that this method can effectively extract the vascular region. (2) A combination method of Hessian matrix and multi-scale analysis was proposed to extract the skeleton of blood vessels. Firstly, the Hessian matrix principle was introduced and the reason to adopt this method was proposed. According to the gray-scale characteristics of blood vessels, an improved combination method of the Hessian matrix and multi-scale analysis was proposed. Hessian matrix was used in the two-dimensional coronary angiography, one of the eigenvectors was in the direction of the vascular axis, and the other is vertical to the direction of the vascular axis. By constructing a vascular characteristic function and selecting more than one scale value for each pixel, we chose the largest characteristic function value as the output value of the characteristic function. Thus, we got the function diagram of the vascular characteristics, so that could we extract the vascular skeleton. The experiments showed that this method can effectively extract the vascular skeleton.(3) A DICOM medical image display and analysis application is established. Adopting Qt4 framework and DICOM 3.0 standard, we used C++programming language and DCMTK to developed DICOM file reading module and the display system. This system can highlight the extracted coronary vascular tree directly on the image, which laid the foundation for three-dimensional visualization, segmentation and registration of DICOM images.