Knowledge Graph Construction And Disease Prediction Research Based On Electronic Medical Records

The hospital receives a large number of patients every day,which results in a large number of Electronic Medical records(EMR),which records the detailed information of the patient’s process from admission examination,inpatient treatment to discharge,and usually contains semi-structured or unstructured images and texts.It is a very valuable resources that be worth mining and utilizing.With the development of artificial intelligence technology,smart healthcare has become an important research in the current medical field.Medical information has successfully completed the transition from paper medical records to electronic medical records,and research based on electronic medical records has also been rapidly developed.Electronic medical records are of great significance in digital medical research.On the one hand,electronic medical records can be used as auxiliary information to support doctors’ clinical decisions and avoid judgment errors based on experience;on the other hand,they can also help to build an online consultation platform by processing patients’ natural language description for disease prediction and treatment.However,due to the variety of information in electronic medical records and the large number of professional terms,including many unstructured information,it is difficult for people to obtain the required information from it,and the unstructured information in electronic medical records is not easy to store.This leads to the underutilization of it.As a semantic network technology,knowledge graph transforms the connections between various things in the real world into entity and relationship descriptions.Its proposal provides an excellent solution for structured storage and full use of electronic medical records.Because most of the existing researches use electronic medical records to extract only named entities or extract entities and relationships separately,the relationship between entities and relationships is split,and there is room to improve the accuracy of extraction.The pre-training model has a good performance in improving the accuracy of downstream tasks of natural language processing.Therefore,this paper proposes an entity relationship joint extraction method based on the pre-training model for Chinese electronic medical records.In addition,the current research based on electronic medical records usually focuses too much on information extraction and ignores the practical results and clinical applications.Therefore,on the basis of entity relationship extraction,this paper conducts the application research of knowledgegraph construction and disease prediction.This paper analyzes the structure and language characteristics of the electronic medical record,and proposes a method of joint extraction of entity relationships based on the pre-trained model.Combined with BIO annotation method + category +relationship,entity relationship is jointly annotated.Based on existing deep learning methods to extract relevant Information.What’s more,it form a triplet into the Neo4 j graph database,build a disease-centric knowledge graph,clearly show the relationship between disease and symptoms,and further optimize the doctor’s work.At the same time,based on the constructed knowledge graph,calculating the influencing factors between diseases and symptoms,assisting doctors in diagnosis.It is helpful to assist doctor’s diagnosis and the research of online inquiry platform.Experiments show that the research of knowledge graph based on electronic medical records is helpful for the extraction of medical knowledge,and has certain auxiliary significance for the development of smart healthcare.