Research On Optic Disc Detection In Fundus Images

Fundus image is commonly used in medical image processing and is an important tool for diagnosing ophthalmic diseases and other systemic diseases such as hypertension and diabetes.As the main physiological structure of the retina,the accurate detection of the optic disc has a lot of help in automatically extracting other anatomical structures in the retina,detecting retinal lesions and identifying other fundus features.Because of retinopathy,non-uniform illumination,and contrast and other factors,accurate detection of the optic disc in the retina image is a challenging task.The detection of the optic disc in the fundus image is divided into two parts: the localization of the optic disc and the segmentation of the optic disc.The localization of the optic disc is to determine the approximate geometric center of the optic disc area,and the segmentation of the optic disc is based on the precise localization of the optic disc,and needs to extract the region of interest and detect the contour of the optic disc.Aiming at the unique structural features of the optic disc in the fundus image,this paper studies the automatic localization and segmentation of the optic disc.This paper presents a new method of optic disc localization.In view of the fact that the disc is a circular or oval region with higher brightness and its size is within a certain range,this paper presents a simple and effective algorithm for the localization of the optic disc center by matching with a concentric circular convolution template with different sizes.We use the fast Fourier transform to compute convolution and image enhancement and vascular segmentation are not necessary in retinal image preprocessing,so it takes less time.We validated the proposed algorithm in four public datasets,with the accuracy rate of 95%,96%,98.88% and 100%.For the fundus image with a resolution of 565×584,the average run time is 1.1s.Based on the localization of the optic disc,this paper presents an optic disc segmentation method based on the parametric active contour model with elliptical constraint.Aiming at the shape feature of the optic disc in the fundus image,this paper adds the ellipse constraint item in the traditional GVF(gradient vector flow)Snake model to detect the contour of the optic disc.The contours are automatically initialized for the shortcomings of manually initializing the disc contours in the traditional method.The mathematical morphology method is used to detect the edge of the optic disc and then the Hough transform is used to detect a circle.The detected circle is used as the initial contour in the GVF Snake model method with elliptical constraint for the segmentation of the optic disc.The method of this paper is tested and verified on the MESSIDOR data set,and the average accuracy rate of 0.83 is obtained,which proves that the algorithm has better segmentation performance.