A Systematic Survey on Diabetic Retinopathy and Deep Learning Approaches

Diabetic retinopathy (DR) is a significant threat to the vision of human beings; it is affected by patients and needs to be detected in the early stage to reduce the severity of vision. In recent research, the DR become interdisciplinary medical and computer science research. In this paper, based on the importance of this research, our research systematically reviewed the deep learning algorithms for automated DR detection, highlighting datasets, model implementations, and challenges. The study explored the Deep Learning approaches in DR management by assessing performance and clinical importance. In this review, our research also included the comparative study of various model outputs based on the different datasets, needful preprocessing techniques, and model refinement for accurate DR detection. We also discussed the integration of advanced models into clinical practice to improve DR management and reduce vision loss associated with diabetes. DR remains a significant global health problem, and advanced detection methods are used to mitigate vision loss. This literature review systematically examines the role of DL algorithms such as EfficientNet and MobileNet in enhancing DR detection. Through rigorous analysis of IDRiD (Indian Diabetic Retinopathy Image Dataset) datasets, model implementations, and challenges, the study synthesizes current research to explain critical directions and understandings. Our research concluded that integrating Deep learning into clinical research holds a guarantee for improving the accuracy and efficiency of DR detection, thereby improving patient results and quality of life. This review delivers valuable wisdom for researchers and clinicians aiming to advance DR detection methodologies and implement innovative solutions for this condition’s early diagnosis and management.