Fundus Image Classification for Detecting Diabetic Retinopathy Using Convolutional Neural Network with Adam Optimizer

Image classification is labeling images into one of several predetermined classes. Using image classification for early detection of eye diseases is essential in improving patient health outcomes by providing faster, more accurate, and more consistent diagnoses. With technology, early detection and timely intervention become more possible, ultimately saving lives and improving patients' quality of life. This research aims to improve the performance of the CNN fundus image classification model for detecting diabetic retinopathy. The dataset consists of 4 labels: Mild Nonproliferative Retinopathy, Moderate Nonproliferative Retinopathy, Severe Nonproliferative Retinopathy, and Proliferative Diabetic Retinopathy. Deep Learning methods based on the Convolutional Neural Network (CNN) model will be used to classify eye retinal fundus images by adding Adaptive Moment Estimation (Adam) for optimizing artificial neural network training. The evaluation results show that adding the Adam optimization technique to the CNN fundus image classification model increases the model performance value significantly, with an accuracy value from 78% to 81%.