周仲瑛教授诊治脑瘤的知识图谱构建与链接预测研究

Research on Construction of Knowledge Graph and Link Prediction for Diagnosis and Treatment of Brain Tumors by TCM Master Zhou Zhongying

为了帮助中医师诊治脑瘤进行临床经验总结与隐含知识挖掘,基于国医大师周仲瑛教授诊治脑瘤的中医医案构建知识图谱并进行链接预测研究。基于Pandas工具从结构化的医案数据中提取实体和关系,使用Py2neo工具和Cypher查询语言构建知识图谱;使用Neo4j图数据库进行存储和可视化展示;构建一个基于名老中医诊治脑瘤医案的真实世界链接预测数据集,并在该数据集上对三类链接预测模型进行实验对比与研究。实验结果表明,对于小型且具有复杂关系的脑瘤医案知识图谱,张量分解模型可以更准确地预测缺失的实体。

To summarize the clinical experience and implicit knowledge of traditional Chinese medicine practitioners in diagnosing and treat⁃ing brain tumors,a knowledge graph was constructed based on medical cases from professor Zhou Zhongying.He is a renowned TCM master who specializes in treating brain tumors.The purpose of this research was to predict links within the knowledge graph.The process involved ex⁃tracting entities and relationships from structured medical case data using Pandas tool.Then,Py2neo tools and Cypher query language were used to build a knowledge graph which was stored and visualized using Neo4j graph database.To compare three types of link prediction models through experiments,a real-world link prediction dataset based on the medical cases of renowned TCM practitioners in treating brain tumors was created.The experimental results showed that for small-scale but complex relationship-based brain tumor medical case knowledge graphs,tensor decomposition models can more accurately predict missing entities.