A deep learning model framework for diabetic retinopathy detection. (9th December 2022)

• Record Type: Journal Article
• Authors: Padmapriya, M.; Pasupathy, S.; Sumathi, R.; Punitha, V.
• Is Part Of: International journal of networking and virtual organisations. Volume 27:Number 2(2022)
• Page Start: 107
• Page End: 124
• Publication Date: 2022-12-09
• Subjects: diabetic retinopathy -- convolutional neural network -- CNN -- machine learning -- deep learning -- ResNet model; Computer networks -- Periodicals; Information networks -- Periodicals; Virtual corporations -- Periodicals; Electronic commerce -- Periodicals; Management information systems -- Periodicals; Management -- Data processing -- Periodicals; 004.65
• Languages: English
• ISSNs: 1470-9503

Abstract

Diabetic retinopathy (DR) is the typical diabetic eye issue and a main reason of blindness around the world. As per the International Diabetes Federation (IDF), the rates of diabetes would rise to 552 million by 2034. Breakthroughs in computer science techniques inclusive of artificial intelligence (AI) and deep learning (DL) have multiplied opportunities for early detection of DR. This indicates that the likelihood of a patient's healing will improve, and the risk of eyesight loss could be minimised in due course. A deep learning model (ResNet) for medical DR detection was examined in this article. The dataset of Asia Pacific Tele-Ophthalmology Society (APTOS) 2019 was used to train and test the DL model. To demonstrate the vitality of the chosen ResNet model, performance measures and testing accuracy like recall, precision, and F1 score were determined. The modified ResNet model attained a testing accuracy of around 84% even with only a few dataset images. Also, training time and computational complexity were reduced with this simpler model.