A Current Review of Generative AI in Medicine: Core Concepts, Applications, and Current Limitations

Area of Science:

• Medical Informatics
• Artificial Intelligence in Healthcare

Background:

• Generative Artificial Intelligence (AI) models create novel synthetic data, differing from discriminative models.
• Key families include diffusion models, Large Language Models (LLMs), and Large Multimodal Models (LMMs).

Purpose of the Study:

• Provide a foundational overview of Generative AI for non-engineering healthcare professionals.
• Clarify current capabilities, applications, and limitations of Generative AI in medicine.

Main Methods:

• Review of Generative AI model families (diffusion, LLMs, LMMs).
• Analysis of current and specialized healthcare applications.
• Discussion of challenges including knowledge gaps, hallucinations, bias, and interpretability.

Main Results:

• Generative AI applications include synthetic image generation, automated documentation, and training simulations.
• Specialized uses involve diagnostic aids, Retrieval-Augmented Generation (RAG), and multi-agent workflows.
• Potential benefits span imaging, documentation, education, and decision support.

Conclusions:

• Generative AI has transformative potential in medicine but faces challenges like data bias and 'hallucinations'.
• Understanding these technologies is crucial for healthcare professionals, though AI is unlikely to replace physicians.
• Navigating regulatory and ethical issues is vital for successful integration.