Artificial Intelligence Tools for Diabetic Retinopathy Screening in Low-Resource Environments: Review of Global Implementation and Challenges

Abstract

Diabetic retinopathy (DR) remains a leading cause of preventable blindness worldwide, with the burden falling disproportionately on populations in low- and middle-income countries (LMICs), where access to timely screening is limited. Advances in artificial intelligence (AI) have enabled the development of automated DR screening tools with diagnostic accuracy comparable to expert graders, offering a potential solution to overcome some of the unique challenges LMICs face. This review examines the performance, global implementation, and unique barriers of AI-based DR screening in low-resource environments, focusing on five leading systems: IDx-DR, EyeArt, Medios AI, Google’s Automated Retinal Disease Assessment (ARDA), and SELENA+. Evidence shows these systems can increase detection rates, reduce referral delays, and improve access to care. However, their integration into healthcare faces unique challenges including data generalizability, infrastructure, regulation, and clinician and patient acceptance. Strategies such as federated learning, offline-capable devices, and easily-comprehensible AI may help overcome these barriers, enabling AI-driven screening to play a critical role in reducing vision loss from DR globally.