Research On Weighted Variational Model Based On Medical Image Segmentation

As an important branch in the field of image processing,selective image segmentation can help doctors find the lesions and cancerous parts of patients earlier in the medical field.In order to achieve the goal of early detection,early treatment and early cure.In this paper,through the study of the patient's lung,liver,chest,knee bone and other parts,and optimize a weighted variational selective image segmentation model for medical images,and named the weighted model based on multi-objective selective image segmentation.In the field of image segmentation,models based on weighted variational method and differential equation are widely favored because of their flexibility and convenient calculation.Variational segmentation models can be divided into edge-based models and region-based models.The edge-based approach utilizes the local properties of the first and second derivatives of the image.The research is carried out in two steps.In the first step,a smooth approximation to the target region related to the Mumford-Shah(MS)model is obtained from the calibrated medical images of the target,and then a weighting function is used to make the approximation provide a larger value for the target region and a smaller value for the other regions.The second step is to use the approximation and perform threshold processing to obtain the image of interest.The main research contents of this paper are as follows:(1)Firstly,the effect of the existing MS active contour model in image segmentation is analyzed,and then the model is optimized for the purpose of multi-objective segmentation.? In order to find the boundary of interest,a convex second-order deformation model is used based on the MS model.?segmentation is obtained by proper threshold value,which is obtained by convex K mean method in this paper.?Distance function d(x)and edge detection function g(x)are defined in the new model.Thirdly,the weighted model based on selective image segmentation is used to segment the image and the weight function is used to adjust the fidelity and smoothness.? In order to improve the operating efficiency,the level set function or membership function is used for numerical calculation;Finally,the existence and uniqueness of U in this study are proved by mathematical proof.In this paper,the experimental results of selective segmentation of medical images are presented:the accuracy of single target recognition is more than 75%,and the accuracy of multi-target recognition is more than 68%,which is 2.5%and 4%higher than MS model,respectively.(2)Medical CT image data sets were made and uploaded,including 7759 healthy liver images,2502 liver cancer cyst images,6749 liver cancer images,500 chest disease images,200 lung disease images,and 300 other images.When marker points were marked near the contour of the target area,it was found that the number and location of marker points had a great influence on the experimental results.The best effect was achieved when the number of marker points was 3-8 and located at the edge of the target area or the inner side of the target area near the edge,with the highest accuracy of more than 90%.The weighted model based on multi-objective selective image segmentation can help medical staff diagnose disease accurately and efficiently and optimize treatment plan.The annotated data set can be used as the basis for analyzing the high incidence data in a certain area and has a wide application prospect.