Translational Medicine-oriented EHR Data Interface And Research Of Diabetes Using Data Mining

With the rise of the Translational Medical as well as the increasing demanding of medical information sharing, building medical information supporting platform base d on electronic health record(EHR) is becoming increasingly important. However, the 1 ack of a unified standard and an efficient integration method between hospitals information system (HIS) results in an incomplete architecture for residents' electronic health record which leads to so-called "isolated information islands". This not only hinders the Translational Medical mode being put into practice, but also results in the inefficient use of the rapid-growing medical big data. As a result, many artificial intelligent related technologies, such as data mining and machine learning, won't perform well on these medical data.This paper studies a two-layer modeled electronic health record standard named "openEHR", then proposed an EHR data interface and designed an openEHR based EHR system based on it. Some key modules are designed and implemented in this paper, including EHR data import, storage and sharing. This system could integrate heterogeneous data into the system while providing a unified EHR data interface for the third party. This builds up a solid foundation for EHR based intelligent medical service.Beyond that, this paper makes researches on diabetes data mining, analysis. Against the lack of quantitative analysis in traditional correlation analysis algorithm, this paper proposed a neural network based sensitivity analysis method. Experiment results show that the algorithm is able to give quantitative risk factors of a given disease. In addition, this paper proposed an improved C4.5 algorithm applied in automatic detection of people at high-risk of diabetes population issue, which is experimentally tested. Lastly, this pater applies all the research above into the development of a chronic health service and analytic system by implementing algorithm modules accordingly. All this work in this paper is of good help and decision support for healthcare managers and doctors.