Research On Automatic Segmentation Of Retinal Blood Vessels

Clinical studies have shown that many important systemic diseases in the human body will cause a specific reaction in fundus.therefore,the fundus image has been widely applied in the medical auxiliary diagnosis field.The morphological structure of retinal vessels is the most susceptible to ocular fundus lesions,the retinal vascular network segmentation and the calculation of arteriovenous width difference contribute to doctors to observe the retinal blood vessels to control of the course of disease.However,manual segmentation blood vessel and determination of width difference requires rich expertise,and is a very tedious and time-consuming task.To this end,the blood vessels automatic segmentation and the change in vessel widths of arteries and veins in retinal images based on computer vision was researched deeply to assist clinical disease investigation and diagnosis.Firstly,aiming at the specificity of retinal blood vessels,the enhancement and the segmentation technology of the fundus image are deeply researched.In order to adapt to different quality image processing needs,the original retinal image is transformed into the color space which is easy to separate the luminance component and to extract the corresponding luminance channel.The each channel of image is enhanced by the method that combining contrast limited adaptive histogram equalization with two-dimensional Gabor wavelet transform so that the contrast of the blood vessel and the background is further improved.The enhanced image is segmented by the maximum interclass variance method and the morphological algorithm is used to post-process the result of image segmentation so as to ensure the effectiveness and accuracy of retinal blood vessel extraction.Secondly,under the premise of accurate segmentation of retinal vessels,a method of measuring the difference between artery and vein width is proposed.In order to reduce the interference of nonvascular areas,round-based Hough transform is used to locate the disc and determine the region of interest.Then,a new 10-dimensional eigenvector based on the centroid of the blood vessel and the pixel of the blood vessel is defined,and the Linear discriminant analysis classifier is adopted to complete the full-automated classification of the artery and vein blood vessel,and then to achieve the calculation of the width of the arteriovenous difference.Finally,DRIVE(Digital Retinal Images for Vessel Extraction)database and clinical fundus data of He's ophthalmology Hospital in Liaoning Province show that the proposed method is accurate in vessel segmentation,artery and vein classification,as well as the calculation of arteriovenous width difference.The proposed method has clinical application and realize value.