Abstract:
This thesis includes research related to analysis of retina to detect and diagnose the Hypertensive Retinopathy (HR), an eye disease, through fundus images. Hypertension can be observed in blood pressure in the form of high blood pressure and it affects the vasculature structure of the human body and retina as well which is mentioned by HR. This disease further leads to other major problems of human body as blindness, heart and brain diseases and strokes, if does not diagnose at earlier stages.
The automatic image analysis very important step to detect and diagnose the disease in short time with less ophthalmologic resources. The advancement in this field has made it easy and possible to detect and diagnose the HR in retinal fundus images at early stages. Before it, it was really difficult and time consuming for the health experts and ophthalmologist to observe and analyze the every image of patient to detect the signs of HR. So, it was an hour of need to develop and assemble such a tool which help the health experts to find out the clinical signs and categories the severity of HR. Initially signs appear in retinal vessels which shows the existence of HR. There are many other changes observed along with vascular changes like papilledema or Optic Disk Swelling. At advance stage of HR, Papilledema can also be observed in retina. That’s why to analyze and detect the main vascular changes along with papilledema, we need some system which calculates the changes in vessels and also detects the optic disk swelling in retina. We can also evaluates the severity and presence of HR by the measurement of Arterio-Venous ratio. And basically this is the ratio of retinal arteries width to veins width. For the automatic detection of HR, CAD (Computer Aided System) has been designed. And this automated system measures the AVR by separating the retinal vessels into two types and detects the grades of papilledema. This automated system is consisted on three main steps; first preprocessing, classification of vessels and calculation of AVR and detection of papilledema, and HR grading depending on arterio-venous ratio and OD swelling. This system helps health specialists for examining and diagnosing of disease. This system obtains 87.27% accuracy on correct vessel classification and 84% on AVR calculation.
