Study On CT Image Segmentation Method Based On Deep Walk And Density Peak Clustering

In recent years,digital image processing technology has been developing rapidly,and new technologies are constantly changing.Conceptually,image processing is a collective term for methods and technologies for digital image processing such as denoising,enhancement,restoration,segmentation,and feature extraction.Nowadays,it has shown its advantages in production,life,and medical,military,and aerospace fields,and plays an important role in extracting key information in digital images.Among them,image segmentation is an important link in digital image processing.Usually,it is the area division of image pixels and the required region of interest.With the increasing maturity and method innovation of image segmentation technology,it has been reflected in different application fields,including face recognition,object detection and medical image processing.There are many branches of current image segmentation techniques,among which clusteringbased image segmentation algorithms are one of them.For clustering-based image segmentation algorithms,the main reasons affecting the segmentation results include the following: one,the clustering effect of the clustering algorithm;two,the digital image preprocessing and feature expression methods.Aiming at these two aspects,this paper proposes an image segmentation method based on deep walking and density peak clustering.This paper proposes a method of image segmentation based on density clustering,which cleverly introduces the concept of topological features of data points by expressing the characteristics of the internal topological relationship of the data,and combines it with the current popular density peak clustering algorithms to improve It solves the problems of sensitive threshold selection and low eigenvalue dimension in traditional image segmentation.Based on the proposed new clustering method,the super-pixel and Lab color space feature expressions are combined to extend the clustering method to the field of CT image segmentation.Based on this,we combined the results of image segmentation with clinical dose prediction to realize the study of breast cancer radiation dose prediction based on deep learning.The specific work is as follows:(1)Combining the Deepwalk algorithm and the density peak clustering algorithm,the clustering algorithm is improved.(2)Extract the color features of the image based on Slic superpixels(map irregular color patches to sample points in the clustering algorithm based on superpixels).(3)Segmentation experiments were performed using breast cancer CT images to verify the accuracy of the algorithm.(4)Using segmented images combined with deep learning to achieve dose prediction for breast cancer radiotherapy.Innovative points of this work:(1)Combining the density peak clustering algorithm with the topology structure,a new clustering algorithm is proposed.The algorithm is applied to the segmentation of CT images.(2)On the basis of breast cancer CT image segmentation,the combination of beam angle and Pix2 pix network is used to predict the dose of radiotherapy,helping the dosimeter to improve the work quality and shorten the work cycle.