Enhancing Diabetic Retinopathy Detection: A Reliable Deep Learning Approach

Diabetic retinopathy (DR) is a complexity associated with diabetes mellitus that specifically targets the human eyes, harming the blood vessels in the retina. Initially, it often goes unnoticed or presents with minor vision disturbances. However, as the condition progresses, it can lead to significant vision impairment or even complete blindness. DR typically arises when blood sugar levels are poorly controlled, making individuals with diabetes mellitus particularly vulnerable to its development. Timely detection prevents irreversible vision loss, underscoring the need for effective screening methods. To address this need, our study proposes a deep learning approach, explicitly utilizing a Connected Convolutional Network to aid in the early detection of diabetic retinopathy. Our model categorizes fundus images into severity levels ranging from No DR to Proliferative DR. In this work, Kaggle utilizes the Aptos 2019 Blindness Detection dataset for training and validation. The methodology encom-passes several key steps, including data collection, preprocessing, and model development. Through our proposed methodology, we have achieved classification accuracy above 90%. The primary intention of the work is to devise an automated and robust system for the timely detection of diabetic retinopathy, thereby enhancing patient outcomes and facilitating early intervention