Knowledge Graph and Large Language Model-Based Clinical Decision Support for Traditional Chinese Medicine External Therapy

Abstract

Traditional Chinese medicine (TCM) external therapy involves heterogeneous clinical knowledge concerning symptoms, treatment indications, intervention methods, practitioner experience, and individualized treatment adjustments, which can be difficult to organize within conventional clinical information systems. This study proposes a knowledge graph and large language model (LLM)-based framework for clinical decision support in TCM external therapy. The framework integrates structured treatment knowledge, clinical observations, symptom–intervention relationships, and selected unstructured clinical narratives into a knowledge-oriented representation, while using an LLM to facilitate contextual retrieval, clinical reasoning, and natural-language interaction. Rather than assuming that LLM-generated recommendations are inherently reliable, the framework emphasizes knowledge grounding, traceable evidence, and clinician-centered interpretation. Particular attention is given to challenges involving semantic ambiguity in TCM terminology, incomplete clinical records, heterogeneous treatment descriptions, knowledge conflicts, and the potential generation of unsupported recommendations by language models. The proposed approach may improve the organization and accessibility of TCM external-therapy knowledge and provide more interpretable decision-support information for individualized intervention. Nevertheless, clinical reliability, generalizability, and safety require further evaluation through expert assessment and prospective clinical validation.