医案系统关键技术研究与实现

发布时间：2018-05-12 11:48

本文选题：中医医案 + 主题模型　；参考：《浙江大学》2017年硕士论文

【摘要】：中医是中国的国粹之一,已经经历了几千年的发展。中医医案作为中医传承的重要载体,体现了中医理、法、方、药的综合运用,蕴含了历代名医丰富的临床诊疗经验,对于中医的学习、研究和发展具有"宣明往范,昭示来学"的作用。然而,医案的文体多样、文白混杂、标准化欠佳等特性,对医案的分类、组织和分析挖掘带来了极大的挑战。同时,中医药领域也缺乏专业的医案知识服务系统。本文以"中国工程科技知识中心"项目的医案系统建设为研究背景,以分析、挖掘、展示医案中隐含的知识为目标,主要关注医案系统关键技术应用研究以及医案系统的设计与实现,主要工作有:1)针对医案类别欠缺的问题,提出了一种基于主题模型的医案分类方法,将医案中的中药、方剂、疾病、症状、证候和治法词汇与非概念词汇区别开。该模型能够发现六类词在每一主题下的关联关系,进而学得更具区分度的文本特征表示,提升了医案分类的准确率。2)为确保每篇医案的独特性,使用结合规则的Simhash算法对医案文本进行去重,同时保证了医案集的丰富性与多样性。3)在对处方进行分析,发现与之相近的经典方剂过程中,为了更好地解析处方,提出基于卷积神经网络的处方识别方法。该方法以句子为分割粒度,从医案中自动提取处方,进而体现医家在治病过程中遣方用药的规律。4)为提高服务数据的精准度,提出了一套众包方案,通过用户提交意见,专家审核的方法对系统中的有误数据进行修正。通过少数服从多数算法和DawidSkene算法对用户意见进行质量控制。5)基于以上研究,设计并实现了医案系统,提供医案搜索、分类浏览、医书阅读、处方分析、医案分析、错误修正等服务,并已上线运行。
[Abstract]:Chinese medicine is one of the quintessence of China, has experienced thousands of years of development. As an important carrier of the inheritance of TCM, TCM medical records embody the comprehensive application of TCM principles, methods, prescriptions and medicines, and contain rich clinical experience of famous doctors in the past dynasties, which has a "clear and clear model" for the study, research and development of TCM. The role of learning. However, the characteristics of medical records, such as diverse style, mixed text and poor standardization, bring great challenges to the classification, organization and analysis of medical records. At the same time, the Chinese medicine field also lacks the specialized medical record knowledge service system. This paper takes the construction of medical record system of "China Engineering Science and Technology knowledge Center" as the research background, and aims at analyzing, excavating and displaying the hidden knowledge in medical records. Focusing on the application of the key technology of medical record system and the design and implementation of medical record system, the main work includes: (1) aiming at the problem of lack of medical record category, a method of classifying medical case based on subject model is put forward, in which traditional Chinese medicine and prescription in medical record are classified. Diseases, symptoms, syndromes and therapeutic vocabulary are distinguished from non-conceptual vocabulary. The model can discover the relationship between the six categories of words under each topic, and then learn a more differentiated text feature representation, which improves the accuracy of medical case classification. 2) in order to ensure the uniqueness of each medical case, The Simhash algorithm combined with rules is used to remove the text of medical records, and at the same time, the richness and diversity of medical records are ensured. 3) in the course of analyzing the prescriptions, we find out that in the process of finding similar classical prescriptions, in order to better analyze the prescriptions, A method of prescription recognition based on convolution neural network is proposed. In order to improve the accuracy of service data, the method takes sentence as segmentation granularity, automatically extracts prescriptions from medical cases, and then reflects the rule of medicine used by doctors in the course of treatment. In order to improve the accuracy of service data, a crowdsourcing scheme is put forward, and the opinions are submitted by users. The method of expert audit corrects the incorrect data in the system. Based on the above research, a medical record system is designed and implemented, which provides medical case search, classification browsing, medical book reading, prescription analysis, medical case analysis, etc. Error correction and other services, and has been online.
【学位授予单位】：浙江大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP311.52;TP391.1

【参考文献】