Document Type : Review Article
Authors
1
Professor of Health Education and Promotion, School of Health, Mashhad University of Medical Sciences, Mashhad, Iran.
2
Electronics Engineer, Mashhad Municipality Fire and Safety Services Organization, Mashhad, Iran.
3
M.S. in Software Engineering, Information and Communication Technology Unit, Mashhad Municipality Fire and Safety Services Organization, Mashhad, Iran.
4
MSc in Health Education and Promotion, Bushehr University of Medical Sciences, Bushehr, Iran.
10.22034/hp.2026.588569.1086
Abstract
Background: Artificial intelligence (AI) is a cornerstone of digital health with transformative potential in diagnosis, medical imaging, and clinical management. Clinical translation is constrained by data quality, interpretability issues, and ethical‑legal gaps. This study maps the global AI landscape and evaluates opportunities, barriers, and the infrastructural readiness of Iran’s healthcare system for AI implementation.
Materials and Methods: We conducted a structured narrative review of PubMed, Scopus, Web of Science, and Google Scholar for English and Persian sources to December 2025. Peer‑reviewed articles, policy reports, and official documents on AI in diagnosis, imaging, health‑systems management, and policymaking were independently screened by two reviewers. Data were synthesized using thematic analysis, and methodological quality was appraised with the SANRA scale.
Results: Five principal AI domains were identified: medical imaging, outcome prediction, personalized medicine, drug discovery, and health‑systems management. Globally, deep learning attains specialist‑level diagnostic performance in multiple tasks, but widespread clinical utility is limited by data heterogeneity, algorithmic bias, black‑box interpretability, and regulatory gaps. In Iran, academic activity and pilot initiatives have risen, yet clinical integration is impeded by fragmented data sources, limited interoperability, absent regulatory and legal frameworks, workforce educational deficits, and weak governance. Deploying imported models without local validation increases risk of performance drift and patient safety concerns. Cross‑cutting needs include improved data quality, interoperable infrastructure, ethical governance, and workforce capacity building.
Conclusion: For safe, effective adoption in Iran, AI should function as "augmented medicine," supporting clinicians. An ecosystem approach—upgrading digital infrastructure, ensuring data interoperability and local validation of models, and establishing ethical‑legal frameworks—combined with interdisciplinary collaboration is required to move from research to responsible clinical practice.
Keywords
