周仲瑛辨治原发性支气管肺癌经验知识图谱的构建及应用

Building Knowledge Graph for Diagnosis and Treatment of Lung Cancer with ZHOU Zhongying’s Clinical Experience

目的构建国医大师周仲瑛辨治原发性支气管肺癌(简称肺癌)经验知识图谱,为名老中医经验传承提供基础。方法以周仲瑛辨治肺癌临床医案为研究对象,结合其辨治肺癌的学术思想和诊疗路径,设计构建知识图谱模式层,并采用Bert-BiLSTM-CRF模型对肺癌医案中症状、舌象、脉象实体进行抽取,形成数据层,基于Neo4j图数据库构建知识图谱,并举例说明知识图谱的应用方法。结果筛选出符合要求的医案446例(576诊次),录入Excel 2016建立国医大师周仲瑛辨治肺癌医案数据库,并进行规范化处理。在Neo4j图数据库中,共创建病机、症状、病位、病理因素、病性、舌象、脉象、中药、治法9个实体标签和病机-症状、病机-舌象、病机-脉象、病机-病理因素、病机-病位、病机-病性、病机-治法、病机-中药、病理因素-中药9种关系类型,包含1281个节点和7791条关系,实现对周仲瑛辨治肺癌知识体系的可视化展示及语义检索功能。结论基于Neo4j图数据库构建周仲瑛辨治肺癌经验知识图谱,可将其辨治肺癌过程中具体的症状表现、治疗原则和用药规律进行客观展示。

Objective To build the knowledge graph for diagnosis and treatment of primary bronchial lung cancer(abbreviated as lung cancer)with the clinical experience of traditional Chinese medicine(TCM)master ZHOU Zhongying,laying a foundation for the subsequent application of knowledge graph in the inheritance of famous TCM physicians’experience.Methods The clinical cases of lung cancer treated by TCM master ZHOU Zhongying were taken,and his academic thought as well as diagnosis and treatment path of lung cancer were summarized,so as to build the model layer of knowledge graph.And the medical entities of symptoms,tongue manifestations and pulse manifestation were extracted by Bert-BiLSTM-CRF to form the data layer of knowledge graph.The knowledge graph was constructed in the Neo4j database,and example was given to illustrate how to use it.Results Totally,446 medical cases(576 visits)that met the inclusion criteria were entered with data normalization,to establish a medical records database of TCM master ZHOU Zhongying in diagnosis and treatment of lung cancer with Excel 2016 software.In the Neo4j database,9 entities labels including pathogenesis,symptoms,disease location,pathological factors,disease nature,tongue manifestations,pulse manifestations,Chinese herbal medicine,and treatment principle,and 9 types of relationships including pathogenesis-symptoms,pathogenesis-tongue manifestations,pathogenesis-pulse manifestations,pathogenesis-pathological factors,pathogenesis-disease location,pathogenesis-disease nature,pathogenesis-treatment principle,and pathological factors-Chinese herbal medicine were created,involving 1281 nodes and 7791 relationships,which had the functions of visual display and semantic retrieval of ZHOU Zhongying’s knowledge system of diagnosis and treatment of lung cancer.Conclusion Knowledge graph based on Neo4j database is an effective tool to objectively display the disease symptoms and visualize ZHOU’s treatment principle and medication rules in treating lung cancer.