DiabEye: an Automated Deep Learning Approach for Early Detection of Diabetic Retinopathy

Diabetic Retinopathy (DR) is among the most common retinopathy diseases caused by diabetes and is a leading factor of vision loss and blindness, especially among diabetics. The prevention of severe complications and better outcomes of patients is highly dependent on timely diagnosis and treatment. This study discusses the creation of DiabEye, a deep learning- based automated diagnostic system for Diabetic Retinopathy aimed at developing AI powered solutions. The research utilizes the DenseNet121 model, where the network was trained to maximize validation accuracy of 73.01%, which turned out to be the highest accuracy that was achieved among all tested configurations. The system is able to accurately stratify and classify DR severity into five categories: none, mild, moderate, severe and proliferative DR by analyzing fundus visual. This allows more accurate decision making by the professionals involved in the healthcare service delivery. DiabEye provides effective and automated solution for early detection of DR while enabling real time predictions and minimizing the need for head-on diagnosis, thus improving clinical efficiency. DiabEye can greatly improve DR screening due to its high accuracy combined with real-time capabilities making it a revolutionary software in the field of ophthalmology and public health. This research contributes to one of the steps forward in enhancing the application of AI in medical diagnostics.