A Robust Approach for the Detection of Diabetic Retinopathy at an Early Stage Using Deep CNN

Abstract

Diabetic retinopathy is one of the major complications that affect the eye of the diabetic patient. This is a condition where high sugar levels in the blood damage the blood vessels in the retina and even lead to partial or whole blindness. There are no mild symptoms at the beginning stage, so it is quite hard to find. But if it is not treated initially it severely impacts the vision and sometimes leads to impaired vision. Most diabetic patients are not aware of this condition, and it leads to Severe vision impairment. Diabetic retinopathy requires early diagnosis; otherwise, it paved the way for various ophthalmic problems. Deep learning is one of the booming methodologies nowadays for predicting and diagnosing diseases because of their algorithms which are based on the structure of the human brain. We can find the result of high accuracy for predicting diabetic retinopathy at the initial stage to prevent the patients from suffering from vision loss using deep learning algorithms. “Good health and well-being” is one of the goals of sustainable development, which include targets like 'Reducing mortality from non-communicable diseases. So, this project uses a highly advanced deep convolutional neural network which concatenates the Xception and ResNet50v2 architecture to train and test the data sets of various diabetic patient's eye images to predict diabetic retinopathy at an early stage and diagnose it initially to prevent the patients from further sufferings.