Interactive Medical Image Segmentation Method Based On Random Walks And Level Set

Medical image segmentation plays an important role in clinical diagnosis and treatment,and provides an important reference for doctors to monitor patients' conditions,make surgical plans and conduct postoperative research.Based on the prior knowledge of cardiac anatomy,this thesis combines the random walk method and level set method to achieve accurate segmentation of right ventricle.The main research contents are as follows.Medical images were preprocessed.Through contrast experiments,piecewise linear transformation and median filtering were selected to enhance and de-noise the input medical images,so as to improve the contrast of the image,filter out the image noise,and provide effective image information for subsequent segmentation.Cardiac presegmentation based on K-means clustering and Hough transform.To address the problem that traditional random walk algorithms require a large number of user interactions,a combination of K-means clustering and Hough transform is used for pre-segmentation of the heart to improve effective seed points for subsequent random walk methods,thus reducing the amount of user interactions and improving the operability and accuracy of the algorithm.An improved random walk algorithm was used for right ventricular segmentation.Aiming at the problem that the traditional random walk algorithm only takes gray information into account when constructing the weight function of the edge,which leads to the missegmentation,this thesis improves the algorithm and adds more rich geodesic distance information.Through comparative experiments,it is proved that the proposed method can better obtain the contour of right ventricle.Level set method was used for right ventricular fine segmentation.Using random walk algorithm for segmentation result as the initial position of level set methods,using a level set method based on the DRLSE model,to right ventricular fine segmentation of publicly available data sets,by using DSC and HD the two evaluation index quantitative assessment of the performance of the algorithm,and compared with other segmentation methods of the same data set were analyzed.A lot of experiments and comparative analysis show that the segmentation accuracy of the method in this thesis is high and the performance of robustness is also better,which is an effective method.