Research And Implementation Of Medical Image Segmentation Algorithm Based On Clustering And Level Set

In recent years,with the rapid development of computer science and medical technology,medical diagnostic instruments with medical image analysis functions have been fully applied in clinical diagnosis and treatment.Today’s clinical diagnosis often uses advanced imaging techniques such as X-ray projection and B-ultrasound to assist in treatment and diagnosis.The generated medical images can provide doctors with basic information on shapes,sizes,locations,etc.The doctor can fully and accurately understand the physiological and pathological changes that occur in the human body and can even predict the development of certain diseases in advance.In the process of clinical diagnosis and treatment,doctors often need to have a detailed view of an organ or part of it,which requires that the generated medical image has a clear boundary and rich content,and can truly reflect the basic information of the tissue and organs.Medical pathology finds that there are differences in the individual structure and irregular arrangement of tissues in the human body.Medical imaging instruments are also affected by external environmental disturbances and other imaging equipment during the imaging process.The above reasons make it difficult for medical diagnostic instruments to obtain a clear boundary,content enriched medical images.Therefore,in-depth exploration into the field of medical image segmentation can promote the development of medical imaging technology and accelerate the development of modern medical treatment.In view of the above background,this paper studies related theories and algorithms of image segmentation.First,due to the high resolution,noise,and inconsistency of the original images produced by medical imaging,consider using clustering to clean and denoise the original images.In the process of clustering,pixels are divided into different clusters,which can not only divide similar pixels into clusters,form boundaries,but also eliminate isolated noise points.Then,a variational level set image segmentation algorithm based on Nystrom method is proposed.The algorithm uses Nystrom’s estimated spectral clustering method to cluster the pixels in the image.By sampling the pixels in the image,the approximate eigenvector and similarity matrix are estimated.Then use the spectral clustering method to cluster the pixels in the image to form a cluster image.Finally,the user’s initial interest area is driven to the edge of the contour by minimizing the region-fitting energy model until the convergence condition is satisfied.Later,the above algorithm was modified for segmenting medical images with weak edges or complex internal contours.Considering the advantages of multi-scale and multi-directional analysis of shear waves in high-dimensional data and their application in edge detection,Shearlet was used to optimize the above algorithms.The algorithm uses the discrete Shearlet transform to detect the edge of the basic image,maximizes the edge information of the retained image,and then uses the level set’s energy-driven function to cause the initial contour to evolve to the image boundary.In order to make the level set function and band The distance between the symbol distance functions maintains regularity,and the energy drive function sets the regularization and penalty terms of the level set.Finally,in order to verify the above algorithm,this paper uses Matlab to build a simulation platform for image segmentation.The platform not only includes the classic image segmentation algorithm,but also includes the algorithm of this paper.The platform can not only read normal natural images,but also read medical format images.Evaluate the performance of different algorithms using the similarity and segmentation time of the final target boundary.Experiments show that the improved algorithm is robust,and the segmentation results maintain a higher degree of similarity than the standard template,which greatly saves computation time.