The Segmentation Of Medical Image Sequences Based On Active Contour Model

With the improvement of computer processing ability and the rapid development of digital image acquisition technology.Image processing technology has also attracted widespread attention.Medical image analysis is an important method to assist doctors in the diagnosis and treatment,and it has also been greatly developed.The main contents of medical image analysis include data acquisition,image segmentation,image registration,3D visualization and so on.Among them,image segmentation plays a very important role in medical image analysis.Image segmentation is an important means of extracting specific tissues and quantitative analysis,and is also an important tool to assist doctors in qualitative analysis.Therefore,image segmentation is very important for improving the recognition of medical images,3D measurement of tissues,and assisting doctors in the diagnosis and treatment of lesions.Medical images have the characteristics of fuzziness and heterogeneity when comparing with common images.For example,the quality(spatial resolution,image contrast,image SNR and inspection conditions)of magnetic resonance images(Magnetic Resonance,MR)often limited by technology.Therefore,the accurate segmentation of medical images is difficult to solve.For the segmentation of medical sequence images,the computational efficiency of image segmentation model is also one of the obstacles that restrict its extensive application.Therefore,this thesis is based on the active contour model of image segmentation.By analyzing the existing active contour model features,some problems existing in the application of medical sequence image segmentation are studied.Based on the analysis of medical sequence images,an active contour model based on the spatial information correlation of sequence images is proposed.First of all,as the organization is similar between the two adjacent slices,the proposed model can effectively find similarity between the edge of tissue slices;secondly,for the initialization of level set function,we select the segmentation results of previous slice as the initial level of the current slice's level set function,thus avoiding the invalid evolution the curve,which greatly reduces the computing time;then,for images polluted by noise,the robustness of the local statistical processing of the source image was utilized in our model as the preprocessing for segmentation;finally,we compare the results among our model and other two models.The experimental results show that our proposed mdoel can segment medical sequence images accurately and efficiently,and has good robustness to noise.