Research On Ultrasound Image Segmentation Based On Graph Cuts And Level Set Method

Thyroid cancer is the cancer of head and neck which is a common andfrequently-occurring disease. Thyroid cancer ranks first in head and neck cancer, increasedincidence of thyroid cancer is the fastest solid cancer. Evolving medical imaging technologyis an important means of diagnosis of benign and malignant thyroid tumors. Two-dimensionalgray-scale ultrasound, doppler ultrasound is the basis for the diagnosis of benign andmalignant thyroid tumors, through a combination of three-dimensional ultrasound,angiography and ultrasound elastography and other new diagnostic imaging technique can getbetter diagnosis results. Therefore, to determine the location accurately and correctclassification of benign and malignant thyroid nodules are necessary.Ultrasound image segmentation has strong pertinence, various algorithms are usuallybased on a specific areas, a specific imaging mode, a particular object of interest. Theseproblems make the ultrasound image segmentation has no uniform standards and universalrules. Which is based on graph theory image segmentation method and level set method forgeometric active contour model based image segmentation has been a hot research field inrecent years.For thyroid ultrasound image segmentation, this paper carried out the following tasks:Firstly, analyze and summarize the characteristics of ultrasound images, and process theultrasound images. By contrast to select more suitable for ultrasound image preprocessingfilter, and prepare for getting more accurate segmentation of thyroid nodules subsequently.Secondly, analysis the graph cut method based on graph theory and level set methodbased on geometric active contour model comprehensively, introduced image segmentationmethod based on graph cuts and the basic principles of CV model level set method, from theperspective of qualitative and quantitative analysis based on the graph cuts and energyminimization function between and briefly discusses the analysis of the CV model basic ideaof the level set method and characteristics.Then, combine graph cut algorithm and CV characteristics of each model,and summarize in Figure cut algorithm model and its CV model their similarities and differences,the CV model to do a series of discrete processing model and graph cuts combine to form anew energy functional, complete energy minimization division.Finally, experiments prove that the improved method on the ultrasound imagesegmentation overcome the shortcomings of level set method that the model need tore-initialize and reduce the number of pending points, reducing the amount of computation,and has high robustness, accurate segmentation and so on.