Research On The Construction And Application Of Knowledge Graph For Chinese Electronic Medical Records

At present,hospitals have stored a large amount of clinical data,such as inpatient medical records that the state requires to be kept for at least 30 years,and outpatient medical records for 15 years.With the development of computer technology,hospital medical records have experienced the leap from paper version to electronic version,and they have also experienced the continuous upgrading of electronic medical record systems,generating a large amount of historical data.However,many of these data are unstructured and cannot be directly utilized,so it is of great theoretical significance and research value to study these electronic medical record data to explore the manifestation and treatment of diseases.This research takes Chinese electronic medical records as the research object and post-structuring of electronic medical records as the main research objective,aiming to provide theoretical basis and technical guarantee to assist clinical diagnosis and treatment and patient services.The research content of this topic mainly includes the following aspects:(1)Construct a knowledge graph of Chinese electronic medical records.Entity identification is first achieved by named entity identification method starting from the data preparation.The identified entities are calculated using the Euclidean distance for entity similarity,so as to normalize the entities.Then,large-scale calculations on the attributes of each entity are performed,and the concepts of co-occurrence times and co-occurrence probability are introduced.Finally,the native knowledge graph storage management tool Neo4 j is selected by comparing the storage methods of knowledge graph and good results are achieved.(2)Build a Trans E model based on probability embedding.In order to provide doctors with the most likely disease indications based on patient symptoms and other information,this thesis improves the distance model to obtain a probability embedding based Trans E model.Based on the constructed knowledge graph,the model is trained to form an embedding vector.Finally,the effectiveness of the proposed method is verified by experiments,which shows that it can predict the disease very well.(3)Design a clinical decision support system based on medical knowledge graph.The system applies the constructed medical knowledge graph to predict possible diseases by acquiring the information of chief complaints in electronic medical records.The patient’s symptom description is combined with the triage system to recommend the optimal outpatient doctor.Finally,the feasibility of the theory and the reliability and stability of the system are proved through module testing,performance testing and functional testing.The thesis has 37 figures,15 tables,and 87 references.