【作者】 吴小蓉；

【导师】 吕旭东；

【作者基本信息】 浙江大学 ， 生物医学工程， 2011， 硕士

【摘要】 临床决策支持系统能依据患者信息进行知识推理,为医护人员开展疾病诊治提供辅助手段,一直是医疗信息化发展的重要方向。临床决策支持系统在临床上应用很大程度上依赖于高效且准确的病人信息获取,传统临床决策支持系统应用时需要通过医生手动输入来获取病人信息,严重阻碍了临床决策支持系统在临床上的应用。在临床决策支持系统不断发展的同时,电子病历在临床上已经取得了成功的应用,但由于目前大部分医疗机构的电子病历间往往存在数据异构性,临床决策支持系统种类繁多信息需求不一致,临床决策支持系统与电子病历系统间存在多对多的数据映射问题,导致临床决策支持系统应用过程中自动信息获取困难。为此本论文开展了下述工作：提出并建立了基础信息模型。该模型对决策支持所需的信息进行抽象,定义了表达各类临床决策支持系统信息需求的基本结构,即定义了那些稳定、不变的概念,包括数据类型、数据结构以及基础信息类。对基础信息模型进行实例化时,根据具体信息需求对基础信息类的属性进行数据结构的动态绑定并定义语义来表达特定临床决策支持系统的具体信息需求,从而能够灵活适应临床决策支持系统信息需求的变更,能够实现将临床决策支持系统与电子病历系统间多对多的数据映射问题分解为一对多的数据映射问题。在此基础上,本文提出了基础信息模型及其实例化模型的关系数据库实现方案。为了解决电子病历与基础信息模型实例化模型间存在的模型异构、.结构异构、语义异构、数据信息缺失等数据异构性问题,本论文提出了一套可动态配置的异构数据转换方法。该方法首先建立实现数据转换操作的医学逻辑操作符,并使用扩展巴科斯范式对由医学逻辑操作符组合而成的规则表达式进行形式化描述以消除异构性,能够解决异构性问题,且能够灵活的适应源电子病历数据库的变化。本论文以面向糖尿病、高血压治疗的临床决策支持系统为例,开发了病人信息获取中间件系统原型。该中间件包含病人信息数据库,数据转换模块,以及信息获取接口模块。本文选取了2家医院的异构电子病历为实验对象,基于本论文开发的信息获取中间件对200份病人数据进行信息获取验证。实验结果数据的有效性证明了本文的方法可以面向各类临床决策支持系统提供自动化的病人信息获取服务,对于解决临床决策支持系统应用过程中的信息获取困难问题有重大意义。更多还原

【Abstract】 Numerous successes have been reported using clinical decision support system to improve the quality of patient care. Such usage of clinical decision support system requires the usage of patient data to generate contextually relevant recommendations. Traditional clinical decision support system requires that doctor entry patient data manually, seriously hinder the application of clinical decision support system at the point of care. So we carry out research on the method of heterogeneous data retrieval from local electronic health records to address the problem of patient data retrieval. Facing the reality of heterogeneity of local electronic health records, the diversity of CDSS, whose basic information requirements are common while detailed demands tend to be variable, requires many-to-many mappings between decision-support systems and electronic health records.To solve this problem, this paper proposes a basic information model that extracts information through defining general data types, data structures and classes needed by CDSS. The instantiation of basic information model, which is realized through binding and defining the restrictive relationships of data types and data structures according to semantic information, generates special information model for designated CDSS. The application of proposed method suits the patient information demands of dissimilar CDSS adaptively. This approach will allow for one-to-many mappings instead of the many-to-many mappings between decision-support systems and electronic health records.In order to overcome the difficulties brought about by the model heterogeneity, structure heterogeneity, semantic heterogeneity and data mismatching between EMR and patient information database, this paper proposed a modifiable method based on dual mode data transformation to transform heterogeneous data by building a set of medical logic operators and formally describing rules generated via combination of medical logic operators using EBNF, which avoids the shortcoming of traditional data transformation methods based on hard code. This paper design and develop a general information acquisition component pertaining to providing automatic patient information collection for CDSS based on the research of information model.This information acquisition component is constituted of a patient information database contains appropriate patient information demanded by CDSS, an interface to acquire general information and the key part, an adapter to transform data between EMR and patient information database. This paper testifies the proposed method by transforming information of 200 patients from heterogeneous EMR of two hospitals to Diabetes Mellitus (DM) and Hypertension diagnostic CDSS and the results prove it to be an effective method that fits for providing various CDSS with automatic patient information acquisition service.更多还原