Weighted Convolutional Neural Network for Detection of Microaneurysms and Exudates in Retinal Images

One frequent diabetes consequence that affects the eyes is diabetic retinopathy. It is the world's leading contributor to blindness. Over 93 million diabetics, according to the World Health Organization (WHO), have diabetic retinopathy in some form. An estimated 7.7 million Americans, or 40% of all diabetics, are thought to be affected with diabetic retinopathy. With age and the duration of diabetes, the prevalence of diabetic retinopathy rises. Most people with type 2 diabetes who have had the disease for more than 10 years and nearly all people with type 1 diabetes who have had it for more than 20 years are predicted to acquire diabetic retinopathy at some point in their lives. Here, Weighted Convolutional Neural Networks (WCNNs), a form of Convolutional Neural Network (CNN), are used to detect Microaneurysm and Exudates by using weights in the convolutional operation to give more priority to particular features or regions in the input data. Early signs of diabetic retinopathy include microaneurysm and exudates. It has been noted that WNN's results outperform those of standard CNN, where sensitivity increase from 87.02(CNN) to 88.51(WNN). Similarly, Specificity, Precision and Accuracy increases from 79.49, 91.09 and 84.81 to 82.56, 92.44 and 86.77 respectively. For early detection of diabetic retinopathy grading is used and related results are also compared.