Research On Construction Of Medical Knowledge Graph For Automatic Disease Diagnosis And Its Application

Today,massive complex healthcare data has been collected from many different resources.Among them,the medical data is very important,containing rich knowledge in medicine such as diseases,drugs,and treatments.This provides useful uses for the smart healthcare.How to model these complex data for supporting the applications on healthcare is a very important and difficult problem.To solve this problem,recent researchers propose knowledge graph.Knowledge graph is a hot topic in big data area,and has become one of the foundation information technologies.At present,the research of knowledge graph in medical field is still in its infancy.In order to improve the accuracy of automatic diagnosis,this dissertation puts forward similar case search and automatic diagnosis algorithm based on knowledge graph,and integrates the data of various data sources,including DBpedia and MIMIC-III critical care data,to construct the medical knowledge base.Use the constructed knowledge graph to carry out the test.The main work we have done is as follows:1.We use different data sources to extract main conceptual entities,entity categories and semantic relationships between entities for constructing the medical knowledge graphs.The data sources include:DBpedia,MIMIC-? critical care dataset,Chinese case dataset,clinical medical dataset,and Orphanet dataset.We adopt inverse maximum matching algorithm to analyze all the datasets and extract the entities.With the extracted entities,we analyze semantic relationships between then by using association rules mining algorithms.Finally,we construct a medical knowledge graph containing patient information,disease information and other knowledge,and visualize it using a powerful tool called Neo4j.2.We use medical knowledge graph to support efficient similar case search and effective automatic disease diagnose.For similar cases search,we first model each case as a graph structure,and adopt the graph similarity search algorithm SEGOS to support efficient case search.To evaluate the effectiveness,we conduct a series of experiments to compare the proposed algorithm with a baseline algorithm.For automatic diagnosis,an extended SEGOS algorithm is developed,and a verification experiment has been conducted to evaluate the algorithm.According to the research in this dissertation,the three items of accuracy rate,recall rate and F value are used as the similar case search evaluation indexes,and the accuracy rate is used as the automatic disease diagnosis evaluation index of the disease.The experimental data show the effectiveness of the use of knowledge graph for similar case search and automatic disease diagnosis.