Advanced CNN-Based Approach for Early Diabetic Retinopathy Detection Using Retinal Images

Diabetic Retinopathy (DR) is a leading cause of blindness among adults, making early detection crucial to prevent severe vision loss and improve patient outcomes. This study focuses on developing an automated diagnostic tool using a Convolutional Neural Network (CNN) to accurately classify DR from retinal images. The dataset comprises 2076 high-resolution retinal images meticulously labelled by medical professionals to indicate the presence (DR) or absence (No DR) of the disease. The dataset is divided into training (2076 images), validation (531 images), and testing (231 images) subsets, ensuring a comprehensive evaluation of the model's performance. The CNN model was trained using advanced data augmentation techniques such as flipping, rotation, scaling, and cropping, along with preprocessing steps like normalisation and resizing to standardise the images. Over 60 epochs, the model was trained to learn the intricate features distinguishing DR from No DR images. The results demonstrate the model's high effectiveness, achieving an overall accuracy of 94%. This robust performance highlights the model's potential to significantly aid healthcare professionals by providing accurate preliminary screenings, thereby facilitating timely diagnosis and personalised treatment plans for patients with Diabetic Retinopathy. This study underscores the importance of integrating machine learning techniques in medical imaging to enhance early detection and management of DR.