基于医案与机器学习信息融合的名中医诊疗糖尿病知识发现方法研究

There are some problems such as no standardized criteria for integrative treatment of diabetes and treatment effect depends on the experience of TCM experts, while the inheritance and utilization of academic experience of TCM experts are lacking of integrity and scientificity, which causing the academic knowledge of famous TCM doctor is difficult to service people widely and extensively. And then, the original discovery method for famous TCM doctor's experience on diagnosis and treatment of diabetes based on the fusion of medical records and machine information learning has been put forward to solve this problem. Academic experience of TCM doctor and professor Li Sai-mei and her team has been converted into a inheritance and available scientific knowledge by this project, and been integrated into the expert system model to achieve the materialized of diagnosis and treatment knowledge of famous TCM doctor, for further research of expert system establishing the scientific basis and also exploring expert system in the new ways for integrative treatment of diabetes. .The main study contents are the standardized expression and knowledge base construction of differentiation and demonstration thinking on diabetes diagnosis and treatment of famous TCM doctors, Knowledge discovery methods based on generation principle of attribute partial order structure diagram, and knowledge discovery methods on TCM prescriptions compatibility group structure of famous TCM doctors, prescriptions dose-effect relationship group structure and Prescription-syndrome group structure. .The preliminary experiment results show that the original knowledge discovery method of this project has effectively data mining for the diagnosis and treatment knowledge on diabetes by famous TCM doctor with Chinese medicine and Western medicine. This new high-tech means for the academic inheritance of famous TCM doctor in modern time and the popularization and application of famous TCM doctor expert system are great significance.

针对中西医结合诊疗糖尿病缺乏规范化标准，治疗效果取决于专家经验，而名老中医及专家学术经验传承和利用尚缺乏完整性和科学性，使得名老中医学术知识难以更广泛服务于民众的现状，提出基于医案与机器学习信息融合原理的现代名中医诊疗糖尿病知识发现原创性方法。本项目将名中医李赛美团队学术经验转化成可传承和可利用的科学知识，并集成到专家系统模型中，实现名中医诊疗知识的实体化，探索中西医结合诊疗糖尿病专家系统的新途径，并为进一步研究专家系统奠定科学基础。主要研究内容有：名中医诊疗糖尿病的辨证论治思维的规范化表达和知识库构建；基于属性偏序结构图生成原理的知识发现方法；名中医的方剂配伍群结构、方剂量效关系群结构、方证群结构知识发现方法等。预实验结果证明：本项目原创性的知识发现方法可有效挖掘名中医的中西医诊疗糖尿病知识。这对于为现代名老中医及专家学术经验传承提供新的高技术手段，普及应用名老中医专家系统有着重大意义。

本项目针对中西医结合诊疗糖尿病缺乏规范化标准，治疗效果取决于专家经验，而名老中医及专家学术经验传承和利用尚缺乏完整性和科学性，使得名老中医学术知识难以更广泛服务于民众的现状，提出基于医案与机器学习信息融合原理的现代名中医诊疗糖尿病知识发现原创性方法。本项目将名中医李赛美团队学术经验转化成可传承和可利用的科学知识，并集成到专家系统模型中，实现名中医诊疗知识的实体化，探索中西医结合诊疗糖尿病专家系统的新途径，并为进一步研究专家系统奠定科学基础。全面采集名中医李赛美教授诊疗糖尿病临床医案2079例，构建专门数据库和知识库。完成了形式背景中关于属性特征和对象特征的完整定义。提出了形式背景属性约简约简方法。解决了从形式背景建立属性偏序结构图的问题。初步解决了属性偏序结构图到概念格过渡的问题，提出了利用属性偏序结构图生成概念格的方法。给出了形式背景中覆盖特征的完整定义。建立了基于覆盖的属性偏序结构图生成方法。基于属性偏序结构理论给出李赛美教授诊疗糖尿病的方剂配伍群结构、方证群结构、中药量效群结构等中医诊疗知识的可视化表示方法和结果。构建了名中医李赛美中西医诊治糖尿病专家系统模型。并开展了基于增量学习原理的知识库动态更新算法的研究。实验结果证明：本项目原创性的知识发现方法可有效挖掘名中医的中西医诊疗糖尿病知识。这对于为现代名老中医及专家学术经验传承提供新的高技术手段，普及应用名老中医专家系统有着重大意义。