Integrating Fine-Tuning and Retrieval-Augmented Generation for Healthcare AI Systems: A Scoping Review.

Area of Science:

• Artificial Intelligence in Medicine
• Clinical Informatics
• Biomedical Natural Language Processing

Background:

• Large language models (LLMs) in healthcare face challenges like hallucinations and static knowledge.
• Fine-tuning (FT) embeds domain reasoning, while retrieval-augmented generation (RAG) provides dynamic knowledge access.
• Hybrid FT + RAG frameworks aim to enhance factual accuracy and clinical reliability of LLMs.

Purpose of the Study:

• To synthesize evidence on hybrid fine-tuning and retrieval-augmented generation frameworks in healthcare AI.
• To evaluate the performance and characteristics of FT + RAG systems in biomedical applications.

Main Methods:

• Scoping review of studies implementing explicit FT + RAG hybrids in healthcare or biomedical tasks.
• Searches conducted across PubMed, IEEE Xplore, Google Scholar, and Embase.
• Data extraction on base models, FT strategies, RAG architectures, applications, and performance outcomes.

Main Results:

• Seven studies met inclusion criteria, consistently showing FT + RAG systems outperform FT-only or RAG-only approaches.
• Parameter-efficient FT methods (e.g., LoRA) and varied RAG implementations were noted.
• Benefits include improved accuracy, reduced hallucination, and increased clinician preference and feasibility.

Conclusions:

• FT + RAG frameworks offer a promising path for clinically grounded healthcare AI.
• These hybrids combine domain-specific reasoning with transparent, up-to-date retrieval.
• Future research should focus on standardized evaluation, workflow integration, and governance for safe deployment.