Certain investigations on image processing techniques for early detection of diabetic retinopathy

Abstract

Diabetic related eye diseases are the most common cause of blindness in the world. This thesis suggests methods for the early detection of Diabetic Retinopathy during mass screening in rural areas. Current methods used by the ophthalmologists are manual examination of the eye and it has lesser accuracy. Density analysis is less efficient which leads to think of the automated analysis. The current methods were studied in consultation with the ophthalmologists. The image processing techniques applied to the retinal images automatically detect the presence of any abnormalities during screening. These methods are easy to use and helpful for the Physicians in the screening of Diabetic retinopathy. The optic disk is the point in the eye where the optic nerves enter the retina. Precise localization of optic disk boundary is an important sub problem, which needs attention in ophthalmic image processing. The Principal Component analysis (PCA) was applied only to the clusters and the output of the PCA was given as input for the propagation through radii method .This saves time and gives good results with higher PSNR value. Exudates are the primary signs of diabetic retinopathy which are the cause for blindness. Exudates can be identified as areas with hard white or yellowish color with varying sizes, shapes and locations. Automated detection of optic disk, exudates, blood vessel extraction and also optic disk and exudates detection through the extraction of blood vessel were done using Matlab software. This thesis gives methods for the early detection of diabetic retinopathy using computer and paves the way for creating awareness during screening. The image obtained from the fundus camera during screening was fed to the computer and it shows whether the image is normal or affected by diabetes. In future, using this approach the microaneurysm and haemorrhages can also be detected by which many abnormalities in the retinal image can be analyzed.