Based On Digital Image Ⅱ Diabetic Retinal Vascular Lesions Detection Research

The development of the society, the ageing of the population, changed people’s life andwork habits, diabetes has become the major disease threat to human life, and the developmenttrend of younger age.Due to diabetes early clinical manifestation is not obvious, onlyconfirmed by auxiliary examination, however, at present there are many malpractices auxiliaryexamination, easy to cause error diagnosis results.After years of clinical experience, the radiusof fundus blood vessel arteriovenous radius ratio changes with the duration of diabetes, isclosely related to the severity and the situation.Retinal vascular lesion detection, this paperresearch question, the use of mathematic development environment to establish retinal imageprocessing algorithms library, the algorithms library contains: retinal image preprocessing,image segmentation, retinal retinal vascular centerline extraction and vascular radiusdifference calculation.Because of the retinal image noise, need of retinal image preprocessing.This article USESthe method of image preprocessing includes three steps: first, the green channel component ofretinal image are extracted, to improve the sharpness of the retinal image;Then gaussianmatched filtering processing, eliminate the noise in the retinal images;Finally to sharpen theretinal image, highlight the image edge information.After the retinal image preprocessing,image segmentation in this paper, the image has carried on the retinal blood vessels, this paperdesign the2D Gabor Wavelet transform method, characterized by how many pixels differentfeatures of blood vessels, set up characteristic parameters to obtain the ideal retinal bloodvessels image segmentation.Get ideal after retinal blood vessels image segmentation, thispaper using the Hessian matrix and the gaussian function to determine the combination ofvascular centerline, and then with a circle of the center line, do inscribed circle in fundusblood vessel image, by Imagej software for retinal fundus blood vessel radius of numericalcalculation, find out the inscribed circle radius, ultimately achieve the goal of retinal vascularlesion detection.Taking liaoning ho ophthalmic hospital eye disease screening subjects of the retinalimage as experiment object in the system, through the analysis of the retinal image aftersegmentation, and on the basis of vascular centerline extraction find blood vessel radiusdifference calculation method, and through the data validation retinal vascular lesionsdetection accuracy of diabetes.Experimental results show that the image segmentationtechnology, this article expounds vascular centerline extraction technology and radius difference algorithm, not only can make the image in the aorta and retinal vein narrated, alsocan find the center of blood vessels, calculate the radius difference, finally achieve the purposeof diagnosis of diabetes.