Analysis Of Carotid Intima-media On Sequential Ultrasound Images

Coronary atherosclerotic heart disease (or simply Coronary Heart Disease, CHD) has become a major reason of Chinese disease mortality. Atherosclerosis causes the formation and rupture of the unstable plaque, the formation and shedding of arterial thrombosis, and also causes vascular occlusion; these three factors have been thought to be the main reason of acute Cardiovascular Disease (CVD) Events. Atherosclerosis is a systemic disease, and the carotid artery has the same risk factor and pathophysiological basis of atherosclerosis as the coronary artery. In this research, by studying the carotid intima-media of health people, we can get the risk assessment of CVD events, and then give proper timely prevention and treatment; this processing is of great clinical significance.Ultrasonography diagnosis is non-invasive and non-ionizing, and has become more and more greatly appreciated as a clinical examination method. This research is about sequential ultrasonic images, we propose a complete solution for carotid ultrasonography diagnosis. The main research work and contribution of this dissertation can be summarized as follows:(1) According to the specific features of carotid ultrasonic images, several classic image segmentation methods are discussed. The method based on active contour model is an effective contour extraction algorithm which combines low level information and high level information and is convenient for manual intervention that is suitable for carotid images processing. Traditional active contour model has several defects, such as being sensitive to initial position and unable to evolution to U-shape. So we use GVF active contour model instead of traditional active contour model, and overcome those defects. According to the implement procedure of the GVF active contour model, we propose a semi-automatic segmentation method, in which the rough contour is manual sketched first, and then evolution is applied. This method improves not only the accuracy, but also the efficiency of the segmentation algorithm.(2) Processing the sequential ultrasound images. We predict the initial position by optical flow method from the final segmentation contour of the previous frame image to avoid error. Because of the large blood flow, successive two frame images may change a lot, which means the carotid shapes change a lot. If we directly treat the final segmentation result of the previous frame image as the initial contour of the present frame image, there may be evolution errors. Optical flow field represents the instant velocity vector of every point in the image, which can be used to predict the initial contour for active contour model. The usage of optical flow field eliminates human error,and makes the algorithm more robust.(3) Extracting the characteristic parameters of carotid artery for further analysis with clinical knowledge. Both carotid intima-media thickness (CIMT) and texture characteristic parameters are extracted in this study. Although in the segmentation procedure, we don't make use of texture information, they are of great significance. Different texture reveals different composition of carotid intima-media complex, which means different probability of causing CVD. Moreover, an analysis of CIMT periodic variation with respect to cardiac cycle has been made combined with electrocardiogram; this gives a more accurate understanding about CIMT. The elastic property, strain and strain rate can be calculated from the CIMT's periodic variation.As mentioned, we improve the segmentation algorithm of ultrasonic sequential carotid images in this paper, and propose a new active contour model based on optical flow predicting initial contour. The experiment result shows that the improved algorithm can extract the precise CIMT,and it could be used as supplementary information in the diagnosis of carotid artery for early preventative treatment.