Generative AI: foundational models. Natural Language Processing (NLP) and LARGE Language Models (LLM)

Area of Science:

• Medical Informatics
• Artificial Intelligence in Medicine
• Clinical Decision Support

Background:

• Generative AI and large language models (LLMs) are emerging technologies with potential applications in healthcare.
• Internists require practical guidance on integrating these tools into daily clinical workflows.

Purpose of the Study:

• To provide internists with a practical overview of integrating generative AI and LLMs into clinical practice.
• To analyze adaptation mechanisms, technical requirements, and opportunities for AI in Internal Medicine.

Main Methods:

• Description of primary adaptation mechanisms: fine-tuning and retrieval-augmented generation (RAG).
• Analysis of technical and organizational requirements: infrastructure, EHR integration, security/privacy (GDPR, EU AI Act).
• Highlighting real-world examples in Internal Medicine for report generation, clinical synthesis, and differential diagnoses.

Main Results:

• Generative AI can enhance clinical decision-making, optimize workflows, and reduce administrative burden.
• Identified limitations include AI bias, hallucinations, and the necessity for human oversight.
• Technical requirements involve infrastructure, EHR integration, and robust security/privacy protocols.

Conclusions:

• Internists can responsibly and efficiently incorporate generative AI tools with proper validation and transparency.
• Prospective validation in real-world settings is recommended for ensuring explainable AI.
• The study empowers internists to leverage AI for improved patient care and practice efficiency.