The Research And Implementation Of Plaque Recognition And Segmentation In Coronary OCT Image

Currently,coronary heart disease is considered a disease with a high fatality rate.Therefore,early diagnosis of coronary heart disease is crucial.Optical Coherence Tomography(OCT)is a common imaging method in coronary arteries,which can clearly display the information of various plaques and the degree of stenosis of coronary arteries.In view of the large difference in plaque shape,high randomness and large amount of speckle noise in coronary OCT images,the following research work is listed in this paper:(1)Aiming at the segmentation of various types of plaque regions in coronary OCT images,a new model based on shallow convolutional neural networks(CNN)and improved random walk was proposed.Firstly,a lightweight network structure was designed to classify the OCT image patches and provide initial seed points for the random walk algorithm.;Secondly,an adaptive random walk algorithm is proposed,which can adjust the gray terms and texture terms in the weight function of the random walk algorithm according to the gray distribution characteristics of different types of plaque regions in the coronary OCT image;Finally,we combined mathematical morphology with random walk to achieve a fully automated segmentation of plaque regions in coronary OCT images.(2)Aiming at the precise segmentation of plaque regions in coronary OCT images,a novel model based on Faster R-CNN,fourth-order partial differential equation(FPDE)and global-local active contour model(GLACM)was proposed.Firstly,Plaque area was detected and classified by Faster R-CNN.Then,a new active contour model is proposed by combining the anisotropic diffusion filtering based on FPDE and level set,and considering global and local information of image gray level.Finally,by taking the rectangular box of plaque area detection results as the initial contour of active contour model,and by minimizing its combined energy functional,plaque area detection and segmentation of coronary OCT images with a large amount of speckle noise can be realized.(3)In order to facilitate the communication between patients and doctors and improve the pathogenicity efficiency of hospitals,this paper designed and implemented a platform forcoronary OCT image analysis and diagnosis.This system is written in Java,Python,Bootstrap and MySQL.The entire system is mainly divided into five parts: role module,data module,front-end module,background module and disease auxiliary diagnosis module and the system diagnosis platform can run stably.