Artificial intelligence for diabetic retinopathy detection: A systematic review

The incidence of diabetic retinopathy (DR) has increased at a rapid pace in recent years all over the world. Diabetic eye illness is identified as one of the most common reasons for vision loss among people. To properly manage DR, there has been immense research and exploration of state-of-the-art methods using artificial intelligence (AI) enabled models. Specifically, AI-empowered models combine multiple machine learning (ML) and deep learning (DL) based algorithms to improve the performance of the developed system architectures that are commercially utilized for the detection of DR disease. However, these models still exhibit several limitations, such as computational complexity, low accuracy in DR stage detection due to class imbalance, more time consumption, and high maintenance cost. To overcome these limits, a more advanced model is required to accurately predict the DR stage in the initial stages. For example, the identification of DR disease in the initial stage helps the ophthalmologist to make an accurate and safe diagnosis, and thereby, eyesight-related issues may be treated more effectively. This study conducted a systematic literature review (SLR) to provide a detailed discussion of the background of diabetic retinopathy, its major causes, challenges faced by ophthalmologists in DR detection, and possible solutions for identifying DR in the initial stage. Also, the SLR provides an in-depth analysis of the existing state-of-the-art techniques and system models used in DR diagnosis based on AI, ML, and recently developed DL-based approaches. Furthermore, this present survey would be helpful for the research community to receive information on the recent approaches used for DR identification along with their significant challenges and limitations.