Artificial Neural Networks in Early Diagnosis and Management of Diabetes Mellitus: A Narrative Review of Models, Biomarkers, and Clinical Integration

Abstract

Diabetes mellitus (DM) is a major global health burden, with increasing prevalence and serious complications. Artificial intelligence (AI), particularly machine learning (ML) and deep learning (DL), has emerged as a powerful tool in diabetes management. While several systematic reviews focus on predictive accuracy, fewer studies integrate technological, clinical, and ethical dimensions into a unified perspective. This article aims to critically examine the role of AI in diabetes care, highlighting its applications in screening, diagnosis, disease progression prediction, and personalized treatment, while also addressing clinical, ethical, and implementation challenges. A narrative review approach was adopted, synthesizing evidence from peer-reviewed journals indexed in Scopus, PubMed, and Web of Science (2014–2024). Studies on AI models—including neural networks, support vector machines, and hybrid approaches—applied to diabetes detection, glucose prediction, and complication management were reviewed. Comparative analysis was conducted with conventional clinical methods to illustrate added value and limitations of AI systems. Existing reviews emphasize algorithmic performance but underexplore real-world integration, clinician acceptance, and patient perspectives. Moreover, gaps remain in transparency, explainability, and equity of AI applications across diverse populations. This review synthesizes technological innovations with clinical implications, offering a balanced perspective for researchers, clinicians, and policymakers. It integrates comparative materials, bridging technical and healthcare narratives to guide future translational research. This article provides a qualitative synthesis of existing literature, offering insights that complement quantitative research. While variability in study designs and publication trends may influence generalizability, the review highlights important directions for future investigation. Research priorities include explainable AI, multi-omics data integration, prospective clinical trials, and cross-cultural validation. Interdisciplinary collaboration is crucial to ensure that AI systems in diabetes care remain ethical, transparent, and patient-centred.