Gap analysis of AI ethical guidelines in the medical sector: a comprehensive review

Counts differ because each database indexes a different set of publications. We treat OpenAlex as the canonical count; Google Scholar is not shown (no API, and crawling it violates its ToS).

Ethical principles are essential to the responsible development and deployment of artificial intelligence (AI) in healthcare. This study examines the explicit lexical representation of predefined ethical themes in a selected corpus of 73 healthcare-AI publications. A nine-theme ethical codebook was developed from institutional guidance and relevant peer-reviewed literature. The study applied NLP-supported full-text analysis, inductive topic modeling, deductive codebook-guided content analysis, theme-based corpus analysis, and paper-level statistical comparison. The findings showed substantial variation in thematic representation. Fairness, accountability, and privacy received the strongest overall representation. Fairness had the highest raw keyword frequency, whereas accountability had the highest median adjusted occurrence rate. Transparency showed moderate representation and was mainly associated with explainability-related discussion. Patient autonomy and sustainability and environmental impact showed the weakest representation, while safety, innovation and progress, and collaboration and partnership received comparatively less attention. Friedman and Holm-adjusted pairwise analyses confirmed significant differences among themes. These findings characterize patterns of ethical-theme representation within the selected healthcare-AI literature corpus. The study proposes measurable indicators to support future healthcare-AI research, governance, and responsible system development.