Diabetic retinopathy detection using deep convolutional neural networks

Diabetic retinopathy is when damage occurs to the retina due to diabetes, which affects up to 80 percent of all patients who have had diabetes for 10 years or more. The expertise and equipment required are often lacking in areas where diabetic retinopathy detection is most needed. Most of the work in the field of diabetic retinopathy has been based on disease detection or manual extraction of features, but this paper aims at automatic diagnosis of the disease into its different stages using deep learning. This paper presents the design and implementation of GPU accelerated deep convolutional neural networks to automatically diagnose and thereby classify high-resolution retinal images into 5 stages of the disease based on severity. The single model accuracy of the convolutional neural networks presented in this paper is 0.386 on a quadratic weighted kappa metric and ensembling of three such similar models resulted in a score of 0.3996.