MultiDx: A Multi-Source Knowledge Integration Framework towards Diagnostic Reasoning

2026-04-27

AI Summary

This work addresses the limitations of current large language models in clinical diagnostic reasoning, which suffer from insufficient domain knowledge and reliance on static knowledge bases that poorly align with real-world clinical reasoning pathways. To overcome these challenges, the authors propose a two-stage dynamic diagnostic reasoning framework. In the first stage, the model integrates heterogeneous medical knowledge from multiple sources—including web search results, SOAP notes, and clinical case repositories—to generate an initial diagnosis accompanied by an interpretable reasoning trace. The second stage refines this output through evidence matching, multi-strategy voting, and a differential diagnosis mechanism to produce a final conclusion. Notably, this approach is the first to explicitly model standard clinical reasoning trajectories, substantially enhancing both interpretability and adaptability. Experimental results demonstrate consistent and significant performance gains over existing baselines on two public medical benchmarks.

Technology Category

• Knowledge Representation and Reasoning: Diagnosis and Abductive Reasoning
• Reasoning under Uncertainty: Other Foundations of Reasoning under Uncertainty
• Cognitive Modeling & Cognitive Systems: Conceptual Inference and Reasoning

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Diagnostic prediction and clinical reasoning are critical tasks in healthcare applications. While Large Language Models (LLMs) have shown strong capabilities in commonsense reasoning, they still struggle with diagnostic reasoning due to limited domain knowledge. Existing approaches often rely on internal model knowledge or static knowledge bases, resulting in knowledge insufficiency and limited adaptability, which hinder their capacity to perform diagnostic reasoning. Moreover, these methods focus solely on the accuracy of final predictions, overlooking alignment with standard clinical reasoning trajectories. To this end, we propose MultiDx, a two-stage diagnostic reasoning framework that performs differential diagnosis by analyzing evidence collected from multiple knowledge sources. Specifically, it first generates suspected diagnoses and reasoning paths by leveraging knowledge from web search, SOAP-formatted case, and clinical case database. Then it integrates multi-perspective evidence through matching, voting, and differential diagnosis to generate the final prediction.~Extensive experiments on two public benchmarks demonstrate the effectiveness of our approach.

Problem

• diagnostic reasoning
• knowledge integration
• clinical reasoning
• large language models
• healthcare applications

Innovation

• multi-source knowledge integration
• diagnostic reasoning
• differential diagnosis
• clinical reasoning trajectory
• large language models