Artificial intelligence in retinal imaging for early Alzheimer's disease detection: A review

Abstract

Background Alzheimer's disease (AD) is a progressive neurodegenerative disorder that necessitates early, accessible, and non-invasive diagnostic methods. Objective This review examines the potential of artificial intelligence (AI)-based retinal imaging as a transformative and scalable tool for early AD detection across the full diagnostic continuum, including the preclinical stage. Methods Following PRISMA guidelines, 63 primary studies were selected from an initial pool of 240 articles retrieved from PubMed, IEEE Xplore, Scopus, Web of Science, and Google Scholar (2017–mid-2025). Advancements in optical coherence tomography (OCT), retinal fundus imaging, and OCT angiography are examined for their capacity to capture structural and vascular biomarkers, including retinal nerve fiber layer thinning and microvascular alterations. AI architectures, including convolutional neural networks, vision transformers, and hybrid models, are evaluated for their accuracy in retinal biomarker analysis. Benchmark datasets, including public and private ones, are assessed for their role in supporting AI-based AD research. Results Key challenges are identified, including data heterogeneity arising from variability in acquisition protocols and demographic representation, as well as computational complexity and limited model interpretability. Emerging approaches—notably multimodal data integration and federated learning—offer promising avenues for enhancing diagnostic accuracy while preserving patient privacy. Conclusions The socioeconomic implications of integrating AI-based retinal imaging into clinical workflows are discussed. By synthesizing recent advancements, unresolved challenges, and future directions, this review underscores the transformative potential of AI-driven oculomics in facilitating early AD diagnosis and improving patient outcomes.