心电信号在线数据知识化研究

【作者】 林泽涛；

【导师】 葛耀峥； 陶国良；

【作者基本信息】 浙江大学 ， 机械电子工程， 2005， 硕士

【摘要】 心脏病是威胁人类生命的主要疾病之一,长期以来,心脏病的研究一直是医学界的重要课题。心电图记录心脏生理电活动,其中蕴涵着丰富的反映心脏节律及其电传导的生理和病理信息,是诊断心脏疾病、评价心脏功能的重要依据之一。微处理器应用于心电图辅助分析,提高了分析的速度,促进了分析算法的发展,但由于心电信号分析的复杂性,在准确性等方面,仍有待于进一步的提高。 本文分析了心电图诊断,其实质就是一个从心电数据中获取心脏状态知识的过程,因此将数据知识化的概念引入心电信号处理中。在此基础上,设计了基于数据知识化的辅助分析算法,将小波变换作为心电特征的检测算法,聚类分析用于分析QRS波群,单独一搏分析和串诊断相结合,对MIT-BIH标准心电数据库中心律失常部分类别能有效识别。 全文共分为六章: 第一章:介绍了心电图的基本知识、传统诊断过程和临床用途,阐述了心电图数据分析的过程和研究现状,并简要介绍了数据知识化的发展和应用,提出了将其引入心电信号分析过程中,最后指出本论文需要研究的内容。 第二章:首先分析心电信号中存在的各类干扰和噪声,并讨论了产生原因与特点,然后对其中工频干扰、基线漂移和高频干扰三类噪声,设计了相应的检测和消除算法,以提高心电信号的信噪比。 第三章:介绍了作为特征检测算法的小波变换基本原理,并阐述了该分析方法所具有的三个特性,接着设计了QRS、P和T波特征的小波变换检测算法,最后用标准心电数据库对检测算法进行了检验。 第四章:对聚类分析的基本知识进行了介绍,然后针对该方法和心电信号的特点,确定了聚类分析的处理对象,接着重点设计了适合于对QRS波群分析的聚类算法,最后用标准心电数据库对聚类算法进行了评估,并分析了聚类结果。 第五章:阐述了心律失常的产生原因,并介绍了主要的心律失常类别,然后分别对诊断分析的单独一搏分析和串诊断进行了介绍,并对部分心律失常类别进行了初步诊断。 第六章:对论文工作进行了总结,并对进一步的工作提出了自己的想法。更多还原

【Abstract】 The heart disease is one of the most prominent disease threatening the life of human being. Study of heart disease has for a long time been an important issue in medicine. Electrocardiograph(ECG) records the activity of physiological electricity about heart, which hides abundant information of physiology and pathology reflecting the rhythm and conduction, and becomes an important basis for diagnosing the diseases and evaluating the function of heart. Microprocessor has been applied to ECG auxiliary diagnosis, which has improved the speed of analysis and promoted the development of analyzing algorithm. For the complexity of ECG, the accuracy of auxiliary diagnosis should be improved.The thesis analyzes the process of ECG diagnosis, which is essentially an acquisition of status knowledge from ECG data, so the concept of data knowledge discovery is introduced to processing the signal of ECG. On that basis, a series of auxiliary analyzing algorithms based on data knowledge discovery is designed, including wavelet transforms as detecting algorithm, clustering analysis being used to analyze QRS complex, combining one beat and final string diagnosis. These algorithms can distinguish part kinds of arrhythmia of the MIT-BIH standard ECG data-base.This thesis consists of six sections as follows:Chapter one firstly introduces the basic knowledge, traditional diagnosis process and clinic application of ECG, then expounds the process and recent development of ECG analysis, and briefly introduces the development and application of data knowledge discovery, which is introduced to processing the signal of ECG. At last, the contents of the thesis are outlined.In second section, firstly all types of interference and noises existing in the signal of ECG are analyzed, and which of the reasons and the characteristics are discussed. Detecting and eliminating algorithms are designed respectively against three kinds of noise, power-line interference, base-line wander and high frequency interference, to improve the signal-to-noise ratio of ECG signal.Chapter three introduces the basic theory of wavelet transforms which is used as the detecting algorithm of the features, and states three characteristics of this method,更多还原