Research On Blood Vessel Segmentation And Optic Disc Localization Of Retina Image

With the rapid improvement of computer hardware and software technology,digital image processing technology combined with computer-aided systems has penetrated into various medical disciplines and fields.The eye is an important visual organ of the human body,and the visual symptoms caused by the lesions in the eye structure can be used as the basis for the diagnosis of certain diseases.The relationship is close,so the analysis and research of fundus images are particularly important.Although the traditional manual segmentation method can obtain high accuracy for the feature extraction of the fundus image,it is difficult to meet the clinical needs due to the time-consuming and subjective factors.To solve this problem,a computer-aided system can be used to detect the fundus image.In this paper,based on the current research status of fundus images,the blood vessel segmentation and optic disc positioning of fundus images are realized based on MATLAB research platform.The specific research contents are as follows:(1)Although the current research on retinal fundus image blood vessel segmentation has achieved certain research results,due to the fineness of the small blood vessels,the gray value contrast between it and the surrounding background is small,and there is still insufficient accuracy for the small blood vessel segmentation results.Therefore,the CLAHE image enhancement algorithm is used to enhance the green channel image in the preprocessing,and the enhanced image and the median filter are used to perform the difference operation to filter out some background noise and retain the details of the blood vessels.In order to futher improve the image quality of blood vessel image preprocessing,on the basis of image preprocessing,the image is first subjected to Gabor wavelet transform at different scales;then the extracted Gabor features are used as input to the generalized linear model for blood vessel segmentation.The algorithm of this paper was tested in the DRIVE fundus image database,and the statistical accuracy was 0.9449,sensitivity 0.9008,specificity 0.9347,and AUC 0.9696.It shows that this algorithm has a good segmentation effect on small blood vessels,and its sensitivity is much higher than other blood vessel segmentation algorithms.It achieves good segmentation results in both lesions and normal fundus images.(2)In the current research on optic disc segmentation of fundus images,due to factors such as retinopathy and uneven illumination when taking pictures,the contrast between the brightness of the optic disc and the background is low,which can lead to inaccurate positioning of the optic disc.Therefore,in the pre-processing stage of this paper,first use the AHE image enhancement algorithm to process the red channel image to improve the contrast between the optic disc and the background and eliminate the vascular interference in the optic disc area,and then process the enhanced image through the morphological opening operation In order to eliminate the interference of the contour of the optic disc and the optic disc to be divided into multiple regions in the subsequent processing;in order to further eliminate the influence of the background brightness and the distribution of blood vessels,the result of the opening operation and the median filter are used to perform the difference operation to obtain the optic disc shadow image.In the optical disc positioning stage,the Otsu method(OTSU)is used to perform threshold segmentation and binarization on the preprocessed image,and the connected region is extracted to obtain the candidate region of the optical disc positioning;by limiting the length and width of the circumscribed rectangle of the connected region,Combined with the feature that the blood vessel density in the optic disc region in the blood vessel segmentation image is larger than other regions,the interference region is further filtered out to obtain the final result of optic disc positioning.In the stage of disc division,the result of disc positioning and the coordinates of the circumscribed rectangle are used as the input for the disc division by Hough transform,and the disc division is realized based on Hough circle detection.The disc segmentation algorithm in this paper was experimentally verified in the DRIVE image library,STARE image library and HRF image library,and the disc segmentation accuracy rates of 100%,96.71%and 97.78%were obtained respectively.Experiments show that this method overcomes the inaccurate segmentation of the optic disc due to the low contrast between the OD and the background,and has the characteristics of good segmentation of the image of the lesion fundus,fast segmentation speed,and good performance.