Classification Of Arteries And Veins Of Retina Vessel And Study On Vessel Fractal Dimension

Fundus color image is playing a more and more important role not only inclinical diagnosis and treatment but also in medical research. After Doctors observethe fundus images of patients, they can diagnose whether patients have been attackedby some eye diseases, diabetic Retinopathy, hypertensive, high blood pressure,diabetes, etc. Studies show that retinal vascular anomalies in fundus images and therisk of developing diabetes, high blood pressure, nervous system disease and severityare closely related. Due to the complexity of retinal blood vessels is difficult to bedirectly observed by doctors, therefore, the introduction of digital image processingtechnology is necessary. Because digital image processing technology can assistdoctors to diagnose.Retinal vascular research mainly divided into three stages: vessel extraction,classification of arteries and veins, vascular fractal dimension research. To realizevascular extraction, we achieve and improve the blood vessel extraction method basedon the Hessian matrix. In order to realize classification of arteries and veins, wepropose and achieve two classification methods, one is based on vascular pixels andthe other one is based on vessel segments. The method based on vessel pixels, firstlyextracted the feathers based on each vascular pixel, and then tag every vascular pixelwith artery or vein label, finally statistical pixel kinds of blood vessels and determinethe proportion of vessel types. But the method based on vessel segments, firstlyextracted the feathers based on each vascular segment, and then and then tag everyvascular segment with artery or vein label, finally according to the structurecharacteristics of vascular correct wrong labels. To calculate the fractal dimension ofretinal vessels, we research and achieve the box counting dimension, the Sandboxdimension, and variation dimension methods to calculate the dimensions of the retinalblood vessels. Then we analyze and compare the obtained results.The vascular extraction accuracy is very important, because it can guarantee theaccuracy of classification of arteries and veins and the results of fractal dimensioncalculation of blood vessels. By studying the difference between arteries and veins,we realized automatic classification of arteries and veins. By researching fractaltheory, we find out the suitable methods to calculate the fractal dimension of retinalvessels. All the methods are tested on the public DRIVE fundus image database, thecorresponding experimental results are obtained.