Artificial Intelligence in Healthcare Diagnostics: A Literature Review

Abstract

Artificial intelligence (AI) has rapidly become a central force in healthcare, particularly in diagnostic medicine, where it promises earlier disease detection, improved accuracy, and more personalized care. This structured narrative review synthesizes recent evidence on AI applications in healthcare diagnostics, focusing on methodological approaches, clinical performance, and the ethical and regulatory challenges that shape real-world adoption. A targeted search of PubMed, Scopus, IEEE Xplore, ScienceDirect, and Google Scholar identified 84 eligible articles published between 2020 and 2025. These studies covered medical imaging, predictive analytics, clinical decision support systems, real-time monitoring, implementation in low- and middle-income countries (LMICs), and cross-cutting issues related to fairness, explainability, and governance. Across imaging and predictive tasks, AI systems frequently achieved diagnostic performance comparable to, or exceeding, that of human experts, while also enhancing workflow efficiency and enabling continuous patient monitoring. However, the review also reveals substantial limitations, including dependence on high-quality and demographically diverse datasets, performance degradation when deployed across different institutions and populations, and persistent algorithmic bias that risks exacerbating health inequities. The black-box nature of many models, gaps in explainable AI (XAI), and fragmented regulatory frameworks further complicate safe and trustworthy deployment in clinical environments. To support more responsible integration of AI diagnostics, this review proposes a 6P framework that emphasizes Performance, Provenance, Population, Privacy, Practice integration, and Policy. These dimensions highlight the conditions under which AI can function as a genuinely supportive tool that augments, rather than replaces, clinician expertise. Overall, AI in diagnostic medicine holds considerable promise, but its benefits will only be realized equitably if technical, ethical, and infrastructure-related challenges are addressed through interdisciplinary collaboration and robust governance.

Ähnliche Arbeiten