Research On Liver Local Lesion Segmentation Techniquein Ct Image

Early detection and treatment to small hepatocellular carcinoma are key to reducing liver cancer mortality rate.Computer-aided diagnosis technology can constantly improve the level of liver cancer diagnosis, help doctors more quickly and accurately make the correct diagnosis. The liver lesions segmentation is the basic premise.The level set model for the Chan-Vese can only handle simple target background image. According to the characteristics of liver CT images, this paper proposes an impruded Chan-Vese level set model to extract liver lesions. Experimental results show that this method can extractt liver lesions better and quickly, which only need a small amount of human intervention.Main tasks are as follows:The liver lesion segmentation method at home and abroad is deeply analysised, which focus on the theory of geometric active contour model, the level set method, and the classical model-Mumford-Shah model. Improved Chan-Vese method will be applied to the segmentation of the liver local disease. Chan-Vese model relies on the global nature of the image so that the final evolution curves stop at the object boundary to achieve target segmentation purposes. Aim at the disadvantage of Chan-Vese model can only deal with simple objectives and background image,the local area including lesions was extracted.To improve the contrast of objective and background and reduce the impact of the gradient boundary of the new structured image,background filling was introduced. Otsu method was used for image pre-segmentation, and then this result act as the initial level set curve evolution, which greatly increased the speed of the segmentation process. Experimental results show that the method can accurately and quickly implemented on the cancer lesion region segmentation.The liver lesions inclue a large number of complex information.The one lesion may show different features,making the target region is on longer a single.The pre-improved Chan-vese model can not be used. multi-phase Chan-Vese model level set segmentation method makes it possible, while also avoiding excessive level set function to bring the coverage area of overlap and vacuum.The improved multi-phase Chan-Vese model level set segmentation was applied to the liver complicated lesions segmentation,which introduced local processing,background filling and Ostu multi-threshold method to improve the speed and accuracy of segmentation Experimental results show that the target region can be well segmented.