build clinical knowledge graphs

Current research and applications of biomedical knowledge graphs (BKGs) lack a systematic survey, particularly regarding domain coverage, task taxonomy, and real-world deployment. Method: This paper introduces the first “Domain–Task–Application” three-dimensional analytical framework to comprehensively synthesize BKG construction methodologies across heterogeneous data sources—including molecular interactions, pharmacological databases, and clinical records—as well as core tasks (knowledge management, retrieval, reasoning, and explainability) and implementation pathways. It integrates techniques spanning knowledge extraction, graph construction, graph embedding, symbolic logical reasoning, and explainable AI, augmented with clinical NLP, drug network modeling, and multimodal alignment capabilities. Contribution/Results: The study yields a structured research landscape, clarifies technological evolution trends and engineering bottlenecks, and establishes a standardized methodology to advance precision medicine, drug discovery, and scientific discovery.

The integration of complex, heterogeneous data and the challenges of semantic reasoning in medicine have significantly hindered advances in clinical decision-making and precision healthcare. This study presents the first systematic review of medical knowledge graph research from dual perspectives—applications (including clinical decision support, disease prediction, health recommendation, precision medicine, and medical question answering) and methodologies (encompassing ontologies, semantic web technologies, deep learning–based information extraction, and neuro-symbolic hybrid modeling). It elucidates the pivotal role of knowledge graphs in enhancing interpretability, semantic consistency, and personalized reasoning. The work traces technological evolution and real-world impact, while identifying critical challenges such as insufficient knowledge coverage, data alignment difficulties, reasoning fragility, and privacy-ethics concerns, thereby charting a path toward safe and effective medical AI systems.

Unstructured clinical text introduces substantial data noise, terminological inconsistency, and logical fragmentation, hindering robust AI deployment in healthcare. To address these challenges, we propose a knowledge graph construction framework integrating SNOMED CT standardized terminology with the Neo4j graph database. Leveraging NLP-driven entity-relation extraction, our method structurally represents clinical concepts—including diseases, symptoms, and medications—and their semantic relationships, enabling multi-hop reasoning and terminological normalization. We further generate a high-quality JSON training dataset from the graph and employ it to fine-tune large language models (LLMs) for diagnostic reasoning. This work constitutes the first implementation of computationally executable SNOMED CT relationship modeling within a graph database, establishing a closed-loop for multi-hop clinical inference. Experimental results demonstrate significant improvements in logical accuracy and interpretability of generated diagnostic pathways, offering a scalable, trustworthy paradigm for AI-assisted clinical decision support systems.

Current clinical knowledge graph construction heavily relies on manual curation and rule-based approaches, struggling to handle the semantic complexity and contextual ambiguity inherent in clinical guidelines and biomedical literature—resulting in low automation and insufficient clinical reliability of structured, interoperable medical indicator knowledge graphs. To address this, we propose a guideline-driven, ontology-guided, retrieval-augmented generation (RAG) and large language model (LLM)-integrated framework for automated knowledge graph construction. Our method synergistically combines domain ontology modeling, dynamic multi-source guideline retrieval, structured schema generation, and expert-in-the-loop validation. It significantly improves the accuracy, scalability, and clinical consistency of knowledge extraction. We experimentally constructed a high-quality knowledge graph covering 200+ core medical indicators and empirically validated its effectiveness across downstream tasks—including intelligent diagnosis and treatment, clinical decision support, and medical question answering.

Existing EHR-driven diagnostic models lack physician-like stepwise reasoning and interpretability. Method: We propose DuaLK, a dual-expert framework comprising (1) an LLM-enhanced diagnostic knowledge graph that integrates structured medical knowledge with semantic relations, and (2) a laboratory-test-guided stepwise pretraining task that explicitly models clinical decision pathways. The method unifies knowledge graph construction, LLM-based semantic alignment, lab-signal-driven proxy tasks, and multi-task diagnostic prediction. Results: On four clinical prediction tasks across two public EHR datasets, DuaLK consistently outperforms state-of-the-art baselines, achieving significant improvements in predictive accuracy (average +3.2% AUC) and reasoning interpretability (86.4% inter-annotator agreement in human evaluation). This work establishes a novel paradigm for knowledge-augmented, clinically grounded AI reasoning.

This work addresses the limitations of existing approaches in constructing oncology knowledge graphs from unstructured clinical text, which often lack effective fact verification and semantic consistency. The authors propose an end-to-end KG-RAG framework that integrates multi-agent prompt engineering, retrieval-augmented generation, and ontology-aligned RDF/OWL semantic modeling to directly extract entities, attributes, and relations. To mitigate hallucination and enhance semantic fidelity, the method incorporates an entropy-based uncertainty scoring mechanism and a multi-LLM consensus strategy. Notably, it enables gold-standard-free, self-supervised continuous refinement. Evaluated on PDAC and BRCA patient cohorts, the resulting knowledge graphs demonstrate high clinical credibility, SPARQL compatibility, and significant improvements over baseline methods in precision, relevance, and ontological compliance.

This study addresses the persistent challenge of efficiently translating biomedical knowledge into actionable outcomes, which is hindered by technical and organizational barriers in data integration. It introduces, for the first time, a systematic engineering paradigm centered on “knowledge assembly,” drawing on software engineering principles of composability and reproducibility, with an emphasis on the construction process rather than static artifacts. The proposed approach integrates key technologies—including identifier mapping, entity disambiguation, schema alignment, evidence provenance, typed namespaces, and standardized exchange formats—to design knowledge infrastructure that supports service composition and reproducible pipelines. Through an analysis of six representative knowledge graph systems, the work identifies eight open engineering challenges, offering both domain-specific guidance and a research agenda for advancing biomedical knowledge graph development through software engineering methodologies.

This work addresses the challenge that current clinical practice guidelines (CPGs), typically represented as free-text documents, are ill-suited for explicitly modeling their underlying decision logic in language model training or retrieval. To overcome this limitation, the study introduces a novel approach that first converts CPGs into executable, programmatic decision structures and then leverages these to generate factual and counterfactual question-answer pairs, thereby constructing structured supervision signals for fine-tuning large medical language models. This enables the models to internalize guideline-driven clinical reasoning rather than merely memorizing surface-level text. Experimental results demonstrate an average relative accuracy improvement of 10.28% across four clinical reasoning benchmarks. Furthermore, clinician evaluations confirm that the model’s generated explanations exhibit significantly higher fidelity, validity, completeness, and clarity compared to baseline methods.

Clinical pathways are often published as unstructured flowcharts whose visual encoding logic is not directly computable by automated systems. This work proposes a five-stage pipeline integrating multi-agent collaboration, deterministic graph algorithms, and compiler verification to automatically transform such flowcharts into verifiable, executable HL7 Clinical Quality Language (CQL) code, deployable as FHIR CDS Hooks services. By innovatively combining large language model agents, typed directed graph construction, semantic computability auditing, and Java-based CQL-to-ELM compiler validation, the approach achieves 100% CQL compilation success and zero terminology hallucinations across five UK NHS cancer pathways. It precisely identifies all non-computable nodes—up to 183 per pathway—preserving compilability through placeholders while uncovering 544 governance issues, all without compromising patient pathway coverage.