Level Set Model And Its Application In Medical Images Segmentation

Image segmentation is a key step in image analysis and an important research topic in image processing and machine vision.Medical image segmentation,as an important application field in image segmentation,is playing an important role in clinical medicine.Medical image segmentation refers to the edges of specific tissues or lesions extracted from medical images.However,the edges of these lesions are often accompanied by weak edges,high noise,and uneven grayscale,which makes it difficult for medical image algorithms to effectively segment.Meanwhile with the advancement of medical technology,the existing medical image types are not limited to the traditional grayscale images,but many new color medical images have appeared.This variety of image types will undoubtedly increase the difficulty of image segmentation.The level set model is one of the most representative methods of the active contour model.Its boundaries are represented by the zero level set of the high-dimensional level set function,and the transformation of the contour is also realized by the evolution of the level set function.An advantage of the level set method is that numerical calculations involving curves and surfaces can be performed on a fixed Cartesian grid without having to parameterize points.Meanwhile the level set method can perform complex topological changes such as merge and split.The emergence of level set algorithm has greatly promoted the development of the field of image segmentation,and also provided high application value in many medical fields.In this paper,we deeply studying the principle of level set mode and exploring the key issues of level set in medical images,including active contour model,curve evolution theory,variational principle,and level set method solving process.The main research contents are as follows:1.Aiming at the problem of symmetric medical image segmentation,in order to solve the shortcomings of the traditional level set method in the commonly used color medical image segmentation,the segmentation effect is poor.Taking the typical tongue segmentation as an example,a bilateral area Constrained symmetry level set segmentation method.Combined with deep learning model for optimization training to improve segmentation accuracy,a corresponding initialization method is also proposed.The maincontents are as follows: combined with the latest RCF model in the deep learning method to obtain the edge probability value,instead of the traditional gradient map;combined with the symmetry characteristics of the medical image,a symmetry detection constraint is constructed to form a new symmetry level set model;Finally,a level set initialization method is proposed to achieve accurate segmentation of color medical images.This method has high segmentation accuracy and robustness for color symmetric medical images,and the extracted edge contours fit the boundaries.2.In order to improve the universality of the algorithm model and improve the problem of segmentation of asymmetric medical images,a BCT-DRLSE model based on bilateral region constraints is proposed.The model first combines the superpixel related theory to divide the pixel area near the boundary of the zero level set into superpixel blocks;secondly,it proposes the pixel distance membership function for grayscale medical images and color medical images,and integrates this membership function into In the bilateral bilateral constraint term,the proposed constraint term is based on the fact that pixels near the boundary of the image have a lower degree of difference than pixels in the non-boundary region.The experimental results show that the segmentation effect of the model in this paper is stable and more universal.3.The initial position of the level set model has a great influence on the segmentation results.When the position is not reasonable,it will not only affect the speed of curve evolution,but also lead to wrong segmentation results.The current medical image segmentation methods make more use of image time domain information,and often ignore the transform domain information of the image.Therefore,this paper proposes a novel initialization method for bispectrum reconstruction.This method first transforms the spatial information of the image by Fourier transform to obtain the phase spectrum of the original image frequency domain;then uses the Fourier transform of the saliency image to obtain the amplitude spectrum of the image;and finally,the inverse Fourier transform realizes the bispectral reconstruction of the image method.This initialization method can realize the initialization of medical images close to the target area,and at the same time,it can achieve accurate segmentation of natural images by incorporating depth information.Finally,this paper summarizes the advantages and disadvantages of the algorithm inthis paper,and proposes future improvement directions and work plans.