A Semantic-based Integration Study For The Electronic Medical Records

An electronic medical record(EMR), refers to the systematized collection of patient and electronically-stored health information in a digital document. It may contain a variety of formats, such as figures, images and videos. In other words, information in EMR documents may be stored as structured、 semi-structured or non-structured forms, and the majority of them are non-structured data such as medical history, admissions records and discharge records. As the national health information technology deepening, medical history which are saved for patient health conditions and the clinical diagnosis processes are increasingly being substituted with hospital information management systems(HIS), picture archiving communication systems(PACS), and clinical information systems(CIS). However, this just makes medical history transformed from paper media to electronic medical records. Data in the medical system still exists as archiving status and information in each medical institutions still exists as isolated islands. Integrating this massive, heterogeneous and multi-source electronic medical records information is the most important and critical step in data sharing and providing personalized treatment options.Medical information has a wide range, large amount of data and diverse data sources. On the other hand, different database vendors have distinct medical data models, which increase the difficulty of integrating this heterogeneous and multi-source medical information. The conventional approaches are using federal database and middleware to integrate relational database information. But in practical applications, some drawbacks have been found in different integration approaches such as: poor data consistency, high complexity, low efficiency and low-accurate query results. The semantic technology which uses knowledge representation language RDF(S), and OWL have a good performance in information integration and inference. In recent years, the semantic technology has shown great advantage, which makes the EMRs integration possible.To solve the above problem, this paper provides a semantic-based information integration framework for EMRs, and on the basis of this framework, we integrate information from medical websites and clinical experts. The main contributions include:Firstly, we provide a semantic-based integration framework for EMRs. After studying structure and content of EMRs, we divide it into two parts: 1) latent, unstructured expert diagnostic knowledge; 2) explicit, structured doctor and patient basic information, and unstructured diagnostic process records. Based on the semantic technology, we establish a unified information integration model for diverse type date.Secondly, a model to integrate latent coronary heart disease diagnostic information from experts are presented. According to principles of constructing ontology, we extract diagnostic knowledge from medical websites and clinical experts, built and maintain coronary heart disease diagnostic ontology. We also build a lightweight body(including symptoms and diseases, diseases and therapy, treatment and drugs) based on this framework.Thirdly, a model to integrate the explicit EMRs information which are basic doctor and patient information and the coronary heart disease diagnostic information are presented. To integrate the basic doctor and patient information which exists in the relational database, we utilize D2 RQ to transform these data to triples. As for the explicit disease diagnostic information, we bulid a model to integrate these date. Then use the ICD codes to combine the above scattering integrated date.In summary, this paper proposes EMRs integration framework based on semantic technology. Based on this framework we build the coronary heart disease semantic ontology to integrate massive, heterogeneous and multi-source electronic medical records and make the information share easier. To make the connection with above ontology, we encode EMRs with ICD codes. We believe that our work in this paper will make a contribution to medical information, provide a useful reference for solving similar problems in the future.