Deep learning for diabetic retinopathy detection and classification Using Long Short-Term Memory (LSTM)

Macular edema, diabetic retinopathy, and other apparent microvascular abnormalities in the human retina are common consequences of diabetes. The primary cause of blindness in the past ten years, diabetic retinopathy (DR), is the subject of this study. Preserving blood glucose control and preventing certain diabetic reactions require immediate or early therapy. Due of its intricacy and depth, DR is difficult to diagnose by laborious manual diagnostic procedures. This paper uses a novel methodology. Instead of manually creating features, we used the Long Short-Term Memory (LSTM) model to extract features for the Region-based Convolutional Neural Network (R-CNN) system. It is an R-CNN-LSTM model with two parts. The proposed R-CNN-LSTM system adds certain ground truth labels to training samples, which enables it to first identify the most informative patches and images before learning basic to complicated retinal properties. Using two independent public datasets, the suggested method investigates the retina vasculature and hard-exudate symptoms of diabetic retinopathy. The recommended illumination normalisation methodology improved the deep learning model and excelled other cutting-edge techniques in terms of accuracy, recall, and precision, according to extensive experimental results.