Retinal Vessel Segmentation And Diameter Measurement Of Arteriovenous Vessel Based On OD Location

Retinal fundus contains abundant human physiological information,and any slight changes in the distribution of fundus network structure,changes in the size and depth of the optic disc,changes in vascular arteriovenous diameter,etc.to some extent reflects the change in human physiology and even some early potential disease risk.As shown by published clinical research,the change in the depth and shape of the optic disk is closely related to glaucoma.The area ratio between optic disk and optic cup has been used in the actual detection of glaucoma;changes in the vascular diameter of arteriovenous veins to some extent also reflects the potential risk of cirrhosis and other cardiovascular and cerebrovascular diseases including diabetes,hypertension and arteries.Based on the current difficult issues in the study of retinal fundus images,combined with the basic issues of vascular detection and feature measurement in eye-retinal images in computer-assisted medical diagnosis,this paper presents a general linear model based on MATLAB in blood vessel segmentation accuracy and speed rapid supervision system.Combining various methods of retinal optic disc location,an optic disk location algorithm based on morphological and Hough transform was proposed.Based on optic disc location,a simple and feasible vascular diameter measurement algorithm was implemented.The specific process is as follows:Firstly,two algorithms of retinal blood vessel segmentation,unsupervised algorithms and supervised learning algorithms are reviewed and described respectively;three different algorithms are used to locate the optic disk,which involves appearance characteristics algorithm,retinal vascular structure algorithm,and fusion appearance characteristics and vascular structure algorithm,are reviewed and described respectively;also the vascular diameter measurement contains many algorithms such as the blood vessel gray distribution characteristics method,the blood vessel center line or edge detection algorithm,etc...,a detailed review of these algorithms is carried out,and advantages and disadvantages of each algorithm and research difficulties are also discussed,and thus new ideas are proposed to achieve a better result.Secondly,the process of extracting blood vessel features based on two-dimensional Gabor filter is studied.The features of the two-dimensional Gabor maximum response filter value under multi-resolution are extracted.Through contrast analysis of each channel of the retinal fundus image,the green channel is selected as the processing image;in the Gabor multi-resolution matching process,the region of interest is bloated in order to reduce the influence of edge noise;and finally,the feature extraction vector is normalized.It provides sufficient data support for the supervised algorithm implementation for fundus vessel segmentation.Thirdly,a simple and fast supervised system based on general linear model and Gabor feature extraction is studied and implemented.During the preprocessing of the algorithm image,in order to strengthen the blood vessel characteristics,coarse blood vessel extraction based on Gabor and top hat and bottom hat transformations based on morphological are superimposed as processing images.The experimental results show that the superimposed blood vessel images can eliminate inconsistent background brightness as well as enhancing the contrast between the blood vessel and the background.The algorithm can quickly save segmentation time and realize the requirements of real-time interactive situations.The experimental results are analyzed and discussed in detail based on the DRIVE fundus database.Then,a fast optic disk location algorithm contains coarse location based on morphological processing and precise location based on Hough circle transformation is proposed and implemented.The algorithm makes full use of the grayscale characteristics and shape features of the optic disk,and eliminate the lesion area and noise area by controlling the length-width ratio limitation and the length-width limitation of the connected area to achieve better location effect on the lesion fundus image;and combine the advantages of the Hough transform to precisely locate optic disk.The precise location of the optic disk is performed,and the experiment proved that the disk was well located.Lastly,a blood vessel diameter measurement algorithm based on multi-directional line detection and optic disk location is proposed and implemented.The semi-automatic measurement of blood vessel AVR is achieved by manual labeling of fundus arteriovenous vessels.The main process includes:the extraction of the interest region based on optic disc;the extraction of the center points of the fundus vessel segments;the statistics of vessel pixel points in the multi-directional line detection.Finally,on the grounds of experiments based on the retinal fundus gallery,it is analyzed and discussed in detail.The experiment proves that this algorithm has reference significance in the application of medical diagnosis.Through various number of experimental results,the supervised system based on the general linear model implemented in this paper can realize rapid segmentation with good segmentation accuracy.It can be used for reference in medical real-time interactive applications.The fast OD location algorithm based on morphological processing has achieved good performance in both the speed and accuracy of location.The vascular diameter measurement algorithm based on multi-directional line detection and the optic disk location has a simple principle and a small error compared with the manual measurement result.It has referential significance in the application of medical diagnosis.