Automated screening system for diabetic retinopathy

The purpose is to develop an automatic computerized screening system to recognize automatically the main components of the retina, an important features of background diabetic retinopathy and classify the normal, abnormal and unknown retinal image. This paper has presented 4 main methods to succeed of retinal diagnosis. Firstly, the retinal images are preprocessed via adaptive, local, contrast enhancement Secondly, the main features of a retinal image were defined as the optic disc, and blood vessels. The optic discs were located by identifying the area with the highest variation in intensity of adjacent pixels. Blood vessels were identified by means of a multilayer perceptron neural network, for which the inputs were derived from a principal component analysis of the image and edge detection of the intensity. Next, the background diabetic retinopathy features are identified. Recursive region growing segmentation algorithms were applied to detect the hard exudates. The haemorrhages and microaneurysms were recognised by detecting all feature similar to the blood vessels and removed the vessels out. Finally, all information is accumulated and diagnosed for diabetic retinopathy or a normal retina. The diabetic retinopathy screening technique has been applied to the 484 normal retina images and 283 images with diabetic retinopathy. The sensitivity and specificity for the computerized screening program to classify the images were corrected 80.21% and 70.66% respectively. The computerized screening system has been developed to classify the normal and abnormalities of retinal images. The development of getting higher performance is in progress.