MMed-RAG: Versatile Multimodal RAG System for Medical Vision Language Models

TL;DR

This paper introduces MMed-RAG, a versatile multimodal retrieval-augmented generation system that significantly improves factual accuracy in medical vision-language models across various medical domains.

Contribution

The paper presents a novel domain-aware retrieval, adaptive context selection, and preference fine-tuning strategy to enhance the generality and reliability of RAG in medical applications.

Findings

Achieves 43.8% improvement in factual accuracy on medical VQA and report generation tasks.

Demonstrates effectiveness across five diverse medical datasets.

Enhances alignment and reduces hallucination in medical vision-language models.

Abstract

Artificial Intelligence (AI) has demonstrated significant potential in healthcare, particularly in disease diagnosis and treatment planning. Recent progress in Medical Large Vision-Language Models (Med-LVLMs) has opened up new possibilities for interactive diagnostic tools. However, these models often suffer from factual hallucination, which can lead to incorrect diagnoses. Fine-tuning and retrieval-augmented generation (RAG) have emerged as methods to address these issues. However, the amount of high-quality data and distribution shifts between training data and deployment data limit the application of fine-tuning methods. Although RAG is lightweight and effective, existing RAG-based approaches are not sufficiently general to different medical domains and can potentially cause misalignment issues, both between modalities and between the model and the ground truth. In this paper, we propose a versatile multimodal RAG system, MMed-RAG, designed to enhance the factuality of Med-LVLMs. Our approach introduces a domain-aware retrieval mechanism, an adaptive retrieved contexts selection method, and a provable RAG-based preference fine-tuning strategy. These innovations make the RAG process sufficiently general and reliable, significantly improving alignment when introducing retrieved contexts. Experimental results across five medical datasets (involving radiology, ophthalmology, pathology) on medical VQA and report generation demonstrate that MMed-RAG can achieve an average improvement of 43.8% in the factual accuracy of Med-LVLMs. Our data and code are available in https://github.com/richard-peng-xia/MMed-RAG.