An Ensemble Learning Approach to Detect the Presence of Diabetic Retinopathy based on Inception Module

In recent times, Diabetic Retinopathy (DR) has become a complication of major degree for diabetic patients, in which the blood vessels present in the eye retina are severely damaged. This in most cases has lead to a loss of vision as observed from records of DR patients, and if left untreated, can cause blindness. As per the estimates of the World Health Organization, DR will impact around 224 million people by 2040. This research study presents a convolutional neural network (CNN) model, followed by employing ensemble learning over various ML algorithms through max voting, for the task of image classification for detecting Diabetic Retinopathy using color fundus images, with the goal of improving the accuracy and efficiency of the diagnostic process. The proposed model is designed to address the challenges of real-world data, such as variability in the appearance of images that belong to the same class and class imbalance. The dataset used for this study is EYEPACS, which is widely used for training and testing models for detecting Diabetic Retinopathy. The Metrics that are used for measuring the performance of the proposed model are accuracy, precision, and F1-S core. This study obtains an accuracy of 95.4 percent.