A systematic literature review of deep learning techniques for identification of ocular abnormalities: Current insights and future directions

Ocular abnormalities such as glaucoma, cataracts, and retinoblastoma pose significant global health challenges due to their con- tribution to vision loss and reduced quality of life. This review investigates the application of machine learning (ML) and deep learning (DL) for the detection and classification of these abnormalities emplying fundus eye images. Adopting the PRISMA methodology, a systematic literature review was conducted on 20 studies published between 2016 and 2023. The analysis highlights the prevalence of convolutional neural networks (CNNs) as the most effective algorithm for diagnosing ocular abnormalities, particularly through fundus image analysis. Diverse ML and DL methodologies, including multimodal approaches, were evaluated for their accuracy, efficiency, and adaptability. Furthermore, this paper investigates the research gaps and based on those proposes a comprehensive framework for future research, emphasizing enhanced datasets, explainable AI, federated learning, and integrated healthcare systems to improve the accuracy, interpretability, and real-world applicability of ocular disease detection models. This review provides valuable insights into current trends and identifies key areas for advancing ocular disease detection technologies.