A HGVF-GAC Model Without Re-initialization Based Carotid Ultrasonic Image Segmentation Algorithm

Cardiovascular and cerebrovascular disease(CCD) is one of principal diseases to do harm to human health.According to the World Health Organization, the deaths caused by cardiovascular diseases (CVD) is one-third of total global deaths each year. Correlational study shows that atherosclerosis (AS) is closely related to CCD and assessment of the status of carotid artery wall plays a very important role on judging in advance risk situation of the symptomless.Ultrasonography, which is one of the most frequently used medical imaging methods for the detection of carotid plaque, is convenient and noninvasive. With the development of high frequency ultrasound, resolution ratio of the ultrasonic image becomes higher and higher, and clinical application of ultrasonography also becomes wider. But this brings another problem: it makes workload of the doctors heavier, finally affects rate of missed disease diagnosis.So scholars bring forward Computer Aided Diagnosis (CAD). A complete CAD system based on grayscale ultrasound image consists of edge extracting, feature extraction and classification.The content of this paper is edge extracting of carotid plaque,in the hope of putting forward a carotid plaque ultrasonic image segmentation algorithm that has the better performance and can used for clinical diagnosis. Main finished research work and contributions of this dissertation are as follows:(1) In view of the problems of net topology and the contour evolving in a wrong direction exist in GVF-Snake active contour model when segmenting carotid ultrasonic images, Paragios puts forward GVF-GAC model. This model takes advantage of geodesic active contour (GAC) model: GVF-GAC model combined with level set method can solve the problem of net topology; the contour of GAC model evolves in the direction of the normal, so the problem of contour evolving in a wrong direction doesn't exist in the GVF-GAC model.But this model has bad transportability, so this paper modifies edge map of this model to make it strictly satisfy the inverse linear relation with edge detection function, and improves some operator of this model, raises improved GVF- GAC model which is more close to GAC model.(2) Improved GVF-GAC model cann't solve the problem of deep concavities, and HGVF- Snake active contour model put forwad by Wang.Y can solve this problem, so this paper adds weak edge detector to HGVF-Snake model, combines it with improved GVF-GAC model and puts forwad improved HGVF-GAC model.The experiment has demonstrated that improved HGVF- GAC model could solve the problems of deep concavities and weak edge detection, and its performance is better than improved GVF-GAC model.(3) In view of the problem of long time consuming and bad efficiency exists in improved HGVF-GAC model, this paper makes some improvements in re-initialization of level set function and model-processing object, and puts forward the improved HGVF-GAC model without re-initialization. This new model saves time of re-initialization in process of curve evolution; moreover, this dissertation furtherly puts forward an improved subimage to be model-processing object, the experiment shows that this method could furthurly decrease program running time, and solve the problems of unstable efficiency and additional error of FPF and TNF'result exists general subimage.To sum up, improved HGVF-GAC model without re-initialization has simple principle and can implement easily, this method not only solves problems of net topology and deep concavities but also furthurly solves problem of weak edge detection. Result of experiment shows that improved HGVF-GAC segmentation algorithm without re-initialization have high segmentation precision, could bring a good result in edge extracting of carotid plaque and its performance is much better than GVF-Snake active contour model.