心电信号自动分析的几种算法研究

发布时间：2018-03-24 11:11

本文选题：波形检测　切入点：心律失常判别　出处：《清华大学》2014年硕士论文

【摘要】：心血管类疾病是人类健康的重大杀手,心电图对这类疾病的诊断有着重要的作用。随着各类心电图机的广泛使用,研究人员对心电信号自动分析算法进行了大量研究。在此基础上,本文针对可以用于实时、远程监护和辅助诊断的远程无线多参数监护系统对心电信号自动分析算法的需求,研究了心电信号自动分析的几种算法,并且设计和实现了系统中的心电信号自动分析模块,在一定程度上弥补了现有分析算法和软件系统的不足。心电信号自动分析技术包括心电信号预处理、波形检测与特征点定位、心律失常判别三个部分。本文在这三个方面均进行了系统的研究,并提出了几种新的自动分析算法,同时设计和实现的心电信号自动分析模块达到了国家检测标准的要求,可以用于几种心律失常的临床自动诊断。本文的主要研究内容与工作成果如下:1.针对心电信号预处理,首先设计了一套基于整系数滤波器的实时预处理算法,具有简单、快速的优点;其次针对其不足,提出了基于中值、均值滤波器和小波变换的心电信号预处理算法,改进了阈值收缩方法。实验表明这是一种性能优良的预处理算法。2.针对波形检测和特征点定位,首先提出了综合滤波器、多阶差分和香农能量变换的实时QRS波群自适应检测算法,降低了QRS波群实时检测的错误率;其次提出了基于小波包等效滤波器的非实时QRS波群检测算法,综合多个滤波器的检测结果,降低了检测错误率,使用MIT-BIH心律失常数据库全体数据测试,错误率仅有0.16%。3.针对心律失常判别,首先使用波形特征如:相邻RR间期差异度、RR间期、QRS波宽、Q波幅值等十个特征判别常见心律失常;其次提出了基于深度置信网络的常见心律失常判别算法,提高了常见心律失常判别的准确率。4.设计并实现了心电信号自动分析模块,根据其功能需求模块,映射为相应的软件模块,并编程实现了这些模块,最后使用MIT-BIH心律失常数据库中的肢体导联第二导联数据验证了模块的功能。
[Abstract]:Cardiovascular diseases are the major killers of human health. Electrocardiogram (ECG) plays an important role in the diagnosis of these diseases. Researchers have done a lot of research on ECG automatic analysis algorithm. On this basis, this paper aims at the demand of ECG automatic analysis algorithm in remote wireless multi-parameter monitoring system, which can be used for real-time, remote monitoring and auxiliary diagnosis. Several algorithms of ECG automatic analysis are studied, and the module of ECG automatic analysis is designed and implemented. To some extent, it makes up for the shortcomings of the existing analysis algorithms and software systems. The ECG automatic analysis technology includes ECG preprocessing, waveform detection and feature point location. Three parts of arrhythmia discrimination. In this paper, the three aspects are systematically studied, and several new automatic analysis algorithms are proposed. At the same time, the design and implementation of ECG automatic analysis module meet the requirements of the national detection standard. It can be used for clinical automatic diagnosis of several arrhythmias. The main contents and results of this paper are as follows: 1. Aiming at ECG signal preprocessing, a set of real-time preprocessing algorithm based on integral coefficient filter is designed, which is simple. Secondly, a ECG signal preprocessing algorithm based on median, mean filter and wavelet transform is proposed. The method of threshold shrinkage is improved. Experimental results show that it is a preprocessing algorithm with good performance. 2. For waveform detection and feature point location, a synthetic filter is proposed. The adaptive detection algorithm of real-time QRS wave group based on multi-order difference and Shannon energy transformation reduces the error rate of real-time detection of QRS wave group. Secondly, a non-real-time QRS wave group detection algorithm based on wavelet packet equivalent filter is proposed. By synthesizing the detection results of multiple filters, the detection error rate is reduced. Using the MIT-BIH arrhythmia database, the test results show that the error rate is only 0.16.3. Firstly, ten characteristics of waveform, such as RR interval difference and QRS wave width and Q wave amplitude, are used to distinguish common arrhythmias, and then an algorithm based on depth confidence network is proposed to distinguish common arrhythmias. The accuracy rate of common arrhythmia discrimination is improved. 4. The ECG automatic analysis module is designed and implemented. According to its functional requirements, the module is mapped to the corresponding software module, and these modules are realized by programming. Finally, the second lead data of limb leads in MIT-BIH arrhythmia database is used to verify the function of the module.
【学位授予单位】：清华大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TN911.7;R54

【相似文献】