Enhanced detection of diabetic retinopathy using a convolutional neural network approach

Abstract Diabetic retinopathy (DR) is one of the most serious problems of diabetes, resulting in retinal damage and blindness. It causes fluid leakage and visual distortion by destroying the retinal tissue’s blood vessels. As a result, it is critical to develop an autonomous DR detection system. Automated image DR assessment systems can diagnose retinopathy in a clinically effective and cost-efficient manner, preventing diabetic-related blindness. This paper proposes the implementation of a diabetic retinopathy detection system using a convolutional neural network (CNN) model. The paper highlighted the preprocessing techniques, dataset balancing, and data augmentation used to enhance image quality and training performance. Additionally, this paper demonstrates the application of the Inception v3 model in training eye retina images and diagnosing the stage of diabetic retinopathy (DR). This system will play a crucial role in the healthcare sector, enabling ophthalmologists working in hospitals to detect DR at an early stage and provide suitable prescriptions.