【作者】 崔晶；

【导师】 路松峰；

【作者基本信息】 华中科技大学 ， 计算机技术， 2017， 硕士

【摘要】 对电子病历进行后结构化处理就是对各种非格式化的文本进行信息和知识提取,然后进行格式化存储,从而进行信息检索和知识发现。传统对病历的处理方法是根据医生的主观经验,肉眼识别出病历中包含的重要信息,这种方式是费时费力的;国外采用医学自然语言处理病历,只能提取到可以映射到标准术语编码的部分信息;国内的研究主要集中在对某一类内容和形式相对比较单一的病历或报告的研究上,以上对电子病历的处理方法无法满足本系统的要求。本文采用文档集定义,模式提取等手段设计并实现了电子病历后结构化系统,该系统可以对各种病历进行信息提取。提出了一种基于模式匹配的电子病历后结构化方法,所做的主要工作如下:首先归纳了电子病历的文本特征,提出先对电子病历进行分类和定义指标,然后根据指标对电子病历处理的方法,针对需要进一步进行提取的内容,提出了基于模式提取和模式匹配的信息提取方法。设计了本系统的总体架构,将本系统分为三个模块:粗粒度提取模块,模式提取模块,细粒度提取模块,并依次阐述了每一模块的功能,然后给出了每一模块的具体处理流程,在粗粒度提取模块中,根据指标对病历信息进行了提取,在模式提取模块中提出了使用分词和词性标注的方式来提取模式和采用DTW算法计算模式相似度的方法,构建了模式库,在细粒度提取模块中根据构建的模式库和模式库中每一条模式的输出规则对病历内容进行了二次提取。设计了系统的类图,然后根据类图对系统进行了实现,并从信息提取准确度和系统运行时间等方面对本系统进行了测评,实验结果表明,本系统可以以较高的准确率提取出病历文本中的信息,为医疗工作提供有力的支持。更多还原

【Abstract】 The post-structured processing of electronic medical records is to extract information and knowledge from all kinds of text and then store them for information retrieval and knowledge discovery.The traditional way of dealing with medical records is that the doctors use their subjective experience to recognize the important information contained in the medical records.This way wastes time and energy.Foreign countries use medical natural language to extract information but this way can only extract information mapped to standard terms.Domestic research focuses on medical records or report which has single type and form content.All of the above processing method of electronic medical records does not meet the requirements of the system.This paper uses the definition of document set,extraction of text pattern to design and implement this structured system for electronic medical records and the system can extract information from various electronic medical records.This paper proposes a methods based on pattern matching to structure electronic medical records and the main works in this thesis as follow: first we summarize the text characteristcs of electronic medical records and put forward a method to structure medical records: classify electronic medical records and define indicators,then use these indicators to structure electronic medical records.We also propose to use pattern extraction and pattern matching to further extract information.We design the architecture of this system.This system can be divided into three modules: coarse-grained extraction module,the pattern extraction module,fine-grained extraction module.Then we introduce the functions and processing procedure of each module.In the coarse-grained extraction module,we use the indicators to extract information;in the pattern extraction module,we propose to use word segmentation and part-of-speech tagging to extract model and using the DTW algorithm to calculate similarity.Then we construct the pattern library.In the fine-grained extraction module,we use the pattern library and the output rule of every pattern to further extractinformation.We design the class diagram of the system and implement this system.Then we assess this system from the aspects such as information extraction accuracy and system running time.The results of experiment show that this system can extract medical record information from the text with high accuracy,so this system can provide strong support for the medical work.更多还原