Detection of Hypertensive Retinopathy from Retinal Images

Abstract

Any acute damage that can cause visual impairment can be termed as retinopathy. The ocular manifestations due to hypertension are hence called as hypertensive retinopathy. During hypertensive retinopathy, the total blood circulation to the retina falls and that may cause damage to vision. As no visual symptoms exist for hypertensive retinopathy, a detailed investigation of fundus images is required. The key factor for the detection is Arterial-Venous ratio (AVR), which usually falls in the range of 0.667 to 0.75 for a healthy retina. Hypertensive retinopathy may cause the ratio to decrease and for the preliminary stage, AVR falls below 0.5. Hence, an automated system for the detection of hypertensive retinopathy has been proposed in this project, which can be integrated to an ophthalmoscope in future in order to investigate the existence of hypertensive retinopathy automatically. The implemented system uses a single Gabor filter for the noise segmentation. A Gabor filter bank consist of 180 filters which differ each other by unit degree has been used to extract blood vessel structures. The extracted blood vessel structure is fed to the support vector machine algorithm, which discriminates the arteries and veins for the calculation of AVR. Hence, altogether the system detects the presence of hypertensive retinopathy and it gives the AVR as its output which helps the user to grade the disease based on Keith-Wagener-Barker grading system.