DoctorRAG: Medical RAG Fusing Knowledge with Patient Analogy through Textual Gradients

2025-05-26

AI Summary

Existing medical RAG systems predominantly rely on structured knowledge bases, overlooking implicit experiential knowledge embedded in similar patient cases—crucial for clinical analogical reasoning. Method: We propose the first clinical decision-oriented RAG framework integrating dual-source information (structured knowledge and real-world patient cases). It introduces a novel patient-case analogical reasoning mechanism; employs concept-tag-guided hybrid retrieval to jointly query knowledge and patient databases; and incorporates a multi-agent Med-TextGrad module enabling knowledge–case dual-constrained generation with text-gradient optimization. The framework further supports multilingual, multi-task fine-tuning. Contribution/Results: Evaluated on a multilingual, multi-task medical benchmark, our method significantly outperforms mainstream RAG baselines, achieving substantial improvements in response accuracy, relevance, and completeness. Moreover, it enables iterative refinement of reasoning outputs.

Technology Category

• Knowledge Representation and Reasoning: Argumentation
• Data Mining & Knowledge Management: Conversational Systems for Recommendation & Retrieval
• Cognitive Modeling & Cognitive Systems: Conceptual Inference and Reasoning

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

Existing medical RAG systems mainly leverage knowledge from medical knowledge bases, neglecting the crucial role of experiential knowledge derived from similar patient cases -- a key component of human clinical reasoning. To bridge this gap, we propose DoctorRAG, a RAG framework that emulates doctor-like reasoning by integrating both explicit clinical knowledge and implicit case-based experience. DoctorRAG enhances retrieval precision by first allocating conceptual tags for queries and knowledge sources, together with a hybrid retrieval mechanism from both relevant knowledge and patient. In addition, a Med-TextGrad module using multi-agent textual gradients is integrated to ensure that the final output adheres to the retrieved knowledge and patient query. Comprehensive experiments on multilingual, multitask datasets demonstrate that DoctorRAG significantly outperforms strong baseline RAG models and gains improvements from iterative refinements. Our approach generates more accurate, relevant, and comprehensive responses, taking a step towards more doctor-like medical reasoning systems.

Problem

• Integrates clinical knowledge and patient case experience
• Enhances retrieval precision with hybrid mechanisms
• Ensures output adherence via multi-agent textual gradients

Innovation

• Integrates clinical knowledge and case-based experience
• Uses hybrid retrieval mechanism for precision
• Incorporates Med-TextGrad for knowledge adherence