Deep Learning for Precision Diagnosis in Diabetic Retinopathy Screening

Diabetic Retinopathy (DR) is a type of eye disease linked with diabetes that can lead to visual impairment and in extreme cases, even blindness can happen if treatment is not taken. Early identification is crucial in managing the disease effectively. Traditional DR screening methods, which mainly depends on manual assessment by ophthalmologists, which are time-consuming and may delay vital treatments. To solve this issue, propose an automated DR detection system that uses cutting-edge technique on deep learning. The proposed system approach utilizes the ResNet-152, a Convolutional Neural Network (CNN) architecture known for its outstanding performance in image classification. A collection of high-resolution fundus images consist of approximately 3662 samples, from Kaggle's The Asia Pacific Tele-Ophthalmology Society 2019 Blindness Detection (APTOS 2019 BD) dataset was used to train and test the model. It comprises five distinct stages of DR (0, 1, 2, 3, and 4) which the model can accurately classify. The system starts by getting high-resolution patient fundus eye images as input. ResNet-152 architecture extracts meaningful features from the given input and the subsequent activation function gives the classification output. The system performed well in DR detection and classification, achieving an accuracy of 96.7%. In summary, the proposed work demonstrates the need for automating DR detection, which effectively reduces the time and effort required for manual screening. With an high accuracy of 96.7%, the system has great potential to enhance DR early detection, ultimately resulting in patient well-being and improved handling of healthcare concerns of people around the world.