Lung-R1: A Knowledge Graph-Guided LLM for Pulmonary Diagnostic Reasoning

2026-06-10 · Source: Artificial Intelligence · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Emerging Technologies & Innovation, AI in Healthcare · Depth: Expert, quick

Summary

Lung-R1 is a novel LungKG-guided large language model designed to improve pulmonary diagnostic reasoning by addressing the Pulmonary Knowledge-to-Diagnosis Gap. This gap highlights the need for patient-specific, relation-aware reasoning over electronic medical record (EMR) evidence, beyond isolated knowledge recall. LungKG, the first structured pulmonary knowledge graph, underpins Lung-R1, featuring 59,038 nodes and 164,308 edges across 15 entity types and 112 relation types. Lung-R1-14B is trained using KG-constrained reasoning-chain construction and KG-guided reinforcement learning. In a 20-system evaluation, it achieved state-of-the-art performance across Choice, Pulmonary-QA, and EMR Diagnosis, scoring 4.3583 in EMR Diagnosis and outperforming the strongest non-Lung-R1 baseline by 0.1476 points. These results validate the effectiveness of KG-guided training for EMR-based pulmonary diagnosis.

Key takeaway

For AI Scientists and Research Scientists developing diagnostic LLMs, integrating structured knowledge graphs like LungKG is crucial for improving patient-specific reasoning. Your models can achieve state-of-the-art performance in complex domains such as pulmonary diagnosis by employing KG-guided training and reinforcement learning. This approach directly addresses the challenge of moving beyond simple knowledge recall to robust, evidence-based diagnostic capabilities.

Key insights

Knowledge graphs can guide LLM training to bridge the gap between pulmonary knowledge and patient-specific diagnostic reasoning.

Principles

• Reliable pulmonary diagnosis requires relation-aware reasoning over EMR evidence.
• Structured knowledge graphs are reusable diagnostic knowledge resources.

Method

Lung-R1 is trained via KG-constrained reasoning-chain construction and KG-guided reinforcement learning built on LungKG.

In practice

• Develop domain-specific knowledge graphs for complex diagnostic tasks.
• Apply KG-guided training to enhance LLM diagnostic accuracy.

Topics

• Pulmonary Diagnosis
• Knowledge Graphs
• Large Language Models
• EMR Diagnosis
• Diagnostic Reasoning
• Reinforcement Learning

Best for: AI Scientist, Research Scientist, Domain Expert

Related on AIssential

• Case-Aware Medical Image Classification with Multimodal Knowledge Graphs and Reliability-Guided Refinement — Medical image classification improves by integrating multimodal knowledge graphs from similar cases with reliability-guided refinement.
• AeroSpectra Sentinel: An Auditable LLM Prompt-Chaining Decision-Support Workflow for Acute Asthma Risk Assessment from Respiratory Sounds and Clinical Signals — An auditable LLM prompt-chaining workflow can enhance acute asthma risk assessment by integrating respiratory sounds and clinical data.
• Knowledge Graphs and Reasoning LLMs for Finding Simple Yet Effective Transcriptomic Perturbation Predictors — Simple K-nearest neighbour models leveraging knowledge graphs, enhanced by RL-trained LLMs, offer competitive transcriptomic perturbation prediction.
• Active Evidence-Seeking and Diagnostic Reasoning in Large Language Models for Clinical Decision Support — LLMs exhibit a substantial performance gap in active, multi-turn clinical diagnosis compared to static, full-context evaluations.
• PathMem: Toward Cognition-Aligned Memory Transformation for Pathology MLLMs — PathMem integrates structured pathology knowledge into MLLMs via a Memory Transformer for improved diagnostic reasoning.
• Language Models as Semantic Teachers: Post-Training Alignment for Medical Audio Understanding — Aligning audio encoders with medical LLMs via a "semantic teacher" paradigm significantly enhances clinical diagnostic performance.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Artificial Intelligence.