面向转化医学的EHR数据接口与糖尿病数据挖掘研究

发布时间：2019-05-27 13:30

【摘要】：随着转化医学模式的兴起,以及人们对医疗信息共享需求的不断增大,构建以电子健康档案(Electronic health record,EHR)为基础的医疗信息支撑平台变得越来越重要。然而目前我国医院信息系统之间缺乏统一的数据标准和一种有效的数据集成机制,居民的电子健康档案没有形成完整的体系,这导致了人们的医疗数据呈现为一个个分散、封闭和无法共享的“信息孤岛”。这不仅仅阻碍了转化医学这种新兴模式的具体实施,也同时造成了急速增长的医疗大数据无法被人们高效利用的尴尬局面。许多当今兴起的数据挖掘、机器学习等人工智能技术在这些医疗数据上的应用效果都会大打折扣。本文针对上述问题,研究了一种采用两层建模架构的开源电子健康档案标准openEHR,设计了一种能够支持多种医疗数据源集成的EHR数据接口,然后以此为基础,初步构建了基于openEHR的健康档案系统,并对系统中关键的数据导入、存储和获取模块进行了设计和实现。该系统可以将不同异构的数据源集成到统一的EHR健康档案系统中,并开放接口提供给第三方使用,这为建立基于EHR的智慧医疗服务提供了很好的支持。在初步实现了EHR数据接口及健康档案系统的基础上,本文对系统中的糖尿病患者EHR数据进行了数据挖掘研究。首先针对传统相关性分析算法在挖掘疾病危险因素中存在的缺乏定量分析的问题,提出了一种基于神经网络敏感性分析的方法,实验结果表明该算法能够定量的给出各种危险因素对疾病的影响程度。接着本文在研究糖尿病高危人群自动识别和筛选的问题中,分析传统C4.5算法存在结点过多的不足之处,提出了一种改进的C4.5算法,并通过实验测试验证了算法的性能。最后本文将以上的研究成果具体应用到慢性病健康服务与分析系统中,设计和实现了相应的算法模块。对于健康管理师和医师实施健康干预起到了很好的辅助建议和决策支持的作用。
[Abstract]:With the rise of transformational medical model and the increasing demand for medical information sharing, it is more and more important to build a medical information support platform based on electronic health file (Electronic health record,EHR). However, at present, there is a lack of unified data standard and an effective data integration mechanism among hospital information systems in our country, and the electronic health files of residents have not formed a complete system, which leads to the dispersion of people's medical data one by one. An isolated island of information that is closed and cannot be shared. This not only hinders the implementation of the new model of transformational medicine, but also results in the embarrassing situation that the rapidly growing medical big data can not be used efficiently by people. Many emerging data mining, machine learning and other artificial intelligence technologies in these medical data applications will be greatly reduced. In order to solve the above problems, this paper studies an open source electronic health file standard openEHR, based on two-tier modeling architecture, and then designs a EHR data interface which can support the integration of various medical data sources, and then based on it. A health file system based on openEHR is constructed, and the key data import, storage and acquisition modules are designed and implemented. The system can integrate different heterogeneous data sources into a unified EHR health file system and open the interface for third parties to use, which provides a good support for the establishment of intelligent medical services based on EHR. Based on the preliminary implementation of EHR data interface and health file system, the EHR data of diabetic patients in the system are studied in this paper. In order to solve the problem of lack of quantitative analysis in mining disease risk factors by traditional correlation analysis algorithm, a method based on neural network sensitivity analysis is proposed. The experimental results show that the algorithm can quantitatively give the influence of various risk factors on the disease. Then, in the study of automatic identification and screening of high risk population of diabetes mellitus, this paper analyzes the shortcomings of the traditional C4.5 algorithm with too many nodes, and proposes an improved C4.5 algorithm, and the performance of the algorithm is verified by experimental tests. Finally, the above research results are applied to the health service and analysis system of chronic diseases, and the corresponding algorithm modules are designed and implemented. It plays a good role in auxiliary advice and decision support for health managers and doctors to carry out health intervention.
【学位授予单位】：浙江大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：R587.1;TP311.13

【相似文献】