Research And Application Of CT Image Weak Edge Detection Method

Computed Tomography(CT),as one of the diagnostic tools of Corona Virus Disease2019(COVID-19),can help doctors observe the lesions of patients.However,there is an edge,which the gray scale change is not obvious and the gray eigenvalue is not much different from the background gray value in the lesion,with a wide transition region,called “weak edge”.In medicine,the weak edge can prevent doctors from accurately calculating the size of a lesion.In order to solve the problem of inaccurate weak edge extraction of lesions,this paper takes CT images of COVID-19 as the research object,proposes two weak edge detection methods,and completes the corresponding system development.(1)Aiming at the problem of low detection accuracy of multi-scale morphology,the GSMM algorithm is proposed,which is combined Gaussian Filter,single-scale Retinex(SSR)and improved multi-scale morphology.Firstly,Gaussian Filter and SSR are used to denoise the image,and then the median filter and the multi-scale Filter is added to the multi-scale morphology for secondary denoising.After the edges are obtained,particle swarm optimization(PSO)is used for binarization.Finally,comparing the GSMM algorithm with other algorithms,the experiment proves that the GSMM algorithm can detect weak edge,but the continuity of weak edge is poor.(2)Aiming at the problems of poor continuity of the GSMM algorithm and easy noise interference of the polynomial interpolation,the PMPI algorithm is proposed,which is combined anisotropic diffusion model(PM)and improved polynomial interpolation.Firstly,the PM algorithm is applied to denoise the image.Then,the polynomial interpolation is used to locate edges,and the area method is uesed to deal with outliers.Finally,the gray moments and Zernike moments are compared with the PMPI algorithm.The experiments show that the Precision,Recall,F-Score,Mean Square Error,Signal Noise Ratio and Running Time of the PMPI algorithm are better than the gray moments and Zernike moments.Compared with the GSMM algorithm,the edge detected by PMPI algorithm is more continuous.According to the experimental data,the Accuracy,Recall,F-Score,Mean Square Error,Signal Noise Ratio and Running Time of GSMM algorithm in stone-paving pattern,single-shot ground glass shadow and multi-ground glass shadow are as follows: 0.8097,0.8551,0.8473,3.6367,6.2804 and15.3851,0.7622,0.8531,0.8058,4.9301,7.0283 and 12.0386,0.7581,0.8459,0.7996,5.2581,8.0365 and 11.0357.Compared with the experimental results of PMPI algorithm,it is found that the values of PMPI algorithm are lower than the values of GSMM algorithm,indicating that GSMM algorithm has more advantages in detecting weak edges of lesion,while PMPI algorithm has better denoising effect.Therefore,the weak edge detection effect of PMPI algorithm is better.(3)Based on Graphical User Interface(GUI)of Matrix&laboratory(MATLAB),we developed a weak edge detection system suitable for the GSMM algorithm and the PMPI algorithm.The detected images can be directly imported into the system,and the corresponding algorithm can be implemented by clicking the buttons in the interface,the process and results of image processing can be displayed on a panel to improve the efficiency of edge detection.There are 90 figures,27 tables and 68 references.