Artificial Intelligence and Mobile Health for Early Detection of Diabetic Complications in Underserved Populations

Abstract

Background: Diabetes is a major health issue, particularly in underserved populations with limited access to healthcare. This paper explores how the combination of artificial intelligence (AI) and mobile health (mHealth) applications can facilitate early detection and management of diabetic complications in these communities. Utilizing AI and mHealth together provides a cost-effective solution to help reduce healthcare gaps in resource-limited areas. Methods: The paper proposes three key ideas: (1) A simple retinal screening pathway using smartphone fundus imaging, analyzed on-device, to detect complications such as retinopathy and neuropathy; (2) a minimum viable dataset (MVD) that includes basic health data and a retinal image for risk assessment; and (3) a negative predictive value (NPV)-first approach to prioritize patients who need immediate care, thereby improving resource allocation. The manuscript also emphasizes the implementation of edge AI, federated learning, offline functionality, and model compression to ensure the system functions effectively in low-resource settings. Conclusions: Finally, it recommends measuring success using metrics such as “time-to-action” and “intervention reach”, ensuring improved health outcomes and offering practical solutions for diabetes care in underserved communities, while providing a model for future healthcare improvements.