Automated detection of diabetic retinopathy through image feature extraction

Diabetes is a disease which is caused due to high blood glucose level in the body. If diabetes is left untreated, vision of the diabetic patient will deteriorate as the disease progresses. Vision deteriorates due to the development of various lesions in eye retina such as microaneurysms, exudates, hemorrhages and cotton wool spots; diabetes at this stage is called Diabetic Retinopathy (DR). Vision remains stable during early stages but as the disease progress and if left untreated it leads to blindness. In this paper, an automated diagnosis of DR using a new approach called Hurst Exponent to determine Fractal Dimension (FD) is presented. Various features like Contrast, Correlation, Energy, Homogeneity, and Entropy are extracted from gray level co-occurrence matrix of image. The statistical analysis of DR and Healthy Retinopathy for various extracted features is presented. The Power Spectrum is obtained for input retinal image, which helps ophthalmologist to quickly diagnose DR on visual basis.