动态心电实时质量评估和特征识别研究

 2022-05-15 10:05

论文总字数:41554字

摘 要

穿戴设备记录的心电波形由于噪声的多样性和个体活动的突发性,信号质量参差不齐,经常出现一段心电波形只包括噪声而没有任何有用临床信息的现象。因此,需要对信号质量进行实时评估和反馈。穿戴式心电信号经过质量评估后,可以滤除临床无用信号,剩余信号可用于后续心电特征提取和参数计算等进一步分析。通过质量评估的心电信号,通常也是包含噪声的,特别是运动时产生的运动伪迹噪声,现有经典QRS检测算法在动态心电信号处理准确度上还需进一步提高,开发更为高效和鲁棒的动态心电QRS检测算法具有重要的临床应用意义。本文针对动态心电实时质量评估和特征识别两个主要问题进行研究,完成了以下工作内容:

(1)总结心电信号质量评估和特征识别研究的现状,以及心电信号的产生原理和导联等基础知识。

(2)提出一种基于多模板匹配和相关系数矩阵等指标的动态心电实时质量评估算法。首先进行心电信号预处理;以Pamp;T和’jqrs’算法共同定位心电QRS波,在此基础上计算其他指标;然后以QRS波为中心建立两组长度不同的心拍模板,通过模板匹配技术得到相关系数矩阵;通过主成分分析方法处理相关系数矩阵得到两个最大贡献率;最后综合考虑最大贡献率、RR间期标准差和QRS波个数等指标实现信号质量三分类:临床有用且信号质量好,临床有用但信号质量差,临床无用信号(纯噪声),并滤除临床无用信号。该算法总Kappa系数为0.735,总准确率为82.3%。

(3)提出一种基于集成经验模态分解和希尔伯特变换的改进QRS检测算法。首先通过集成经验模态分解得到固有模态函数,选取其中三个信号加和得到重组信号;其次通过求导、希尔伯特变换,得到希尔伯特包络线,最后设定阈值,通过寻峰算法确定QRS波位置。该算法能有效识别定位动态心电的QRS波,并对于节律异常或噪声信号有一定鲁棒性。在MIT-BIH心律失常标准数据库仿真中,该算法的灵敏度、阳性预测值和准确率分别为96.84%、99.53%和96.40%。

关键词:穿戴式心电,心电质量评估,模板匹配,QRS检测,集成经验模态分解,希尔伯特变换

ABSTRACT

The electrocardiogram(ECG) recorded by the wearable device usually shows different signal quality due to the diversity of noise and individual activities. It is often found that the ECG segment only includes noise without any clinically useful information. Therefore, real-time evaluation and feedback of signal quality are required. After the quality evaluation, wearable ECGs that are not clinically available are filtered, and the remaining signals can be used for further analysis such as subsequent ECG feature extraction and parameter calculation. However, these signals contain noise, especially the motion artifact noise. The existing classical QRS detection algorithm needs to be further improved in the accuracy of dynamic ECGs processing. So it is of great clinical application significance to develop a more efficient and robust dynamic ECG QRS detection algorithm. In this paper, dynamic ECG real-time quality evaluation and feature recognition are studied, and the following tasks are completed:

(1) Summarize the status quo of the research on the quality evaluation and feature recognition of ECGs, as well as the basic knowledge of the generation principle and leads of ECGs.

(2) A real-time dynamic ECG quality evaluation algorithm based on multi-template matching and correlation coefficient matrix is proposed. Firstly, pretreatment of ECGs is performed. Pamp;T and 'jqrs' algorithms are used to locate the QRS waves,and other indicators are calculated on this basis. Then, two heart beat templates of different length are built with the center of QRS waves, and the correlation coefficient matrix is obtained by template matching. Two maximum contribution rates are obtained by PCA. Finally, parameters such as maximum contribution rates, RR interval standard deviation and number of QRS waves are considered comprehensively to realize the three classification of signal quality: clinically useful and good signal quality, clinically useful but poor signal quality, and clinically useless signal (pure noise), and the clinical useless signals are filtered out. The total Kappa coefficient of this algorithm is 0.735, and the total accuracy is 82.3%.

(3) An improved QRS detection algorithm based on Ensemble Empirical Mode Decomposition (EEMD) and Hilbert transform is presented. Firstly, EEMD is used to obtain the IMFs, and three IMFs are selected to obtain the recombinant signal. Secondly, Hilbert envelope is obtained through derivation and Hilbert transformation successively. Finally, the threshold value is set and the peak finding algorithm is used to detect the QRS waves. This algorithm can effectively identify and locate the QRS waves of dynamic ECGs, and has certain robustness against abnormal rhythmic or noisy signals. The sensitivity, positive predictive value and accuracy of the algorithm are 96.84%, 99.53% and 96.40% respectively in the simulation of MIT-BIH arrhythmia database.

KEY WORDS: wearable ECG, ECG quality evaluation, template matching, QRS detection, EEMD, Hilbert transform

目 录

摘 要 I

ABSTRACT II

第一章 绪论 1

1.1研究背景与意义 1

1.2研究现状 2

1.2.1动态心电信号实时质量评估 2

1.2.2特征识别QRS检测算法 3

1.3实验数据库 3

1.3.1质量评估算法数据库简介 3

1.3.2QRS检测算法数据库简介 4

1.4论文内容与撰写安排 4

第二章 心电基础知识 5

2.1心电信号的产生原理 5

2.2心电图的标准导联 6

2.2.1肢体导联 6

2.2.2胸前导联 7

2.3本章小结 8

第三章 动态心电实时质量评估算法 9

3.1ECG波形预处理 10

3.1.1异常信号检测 10

3.1.2基线漂移滤除 11

3.2QRS检测 12

3.3其他指标 13

3.4信号模板匹配 14

3.5相关系数矩阵处理 16

3.6融合评估算法 17

3.7实验结果与分析 17

3.8本章小结 19

第四章 ECG特征识别研究 20

4.1常见QRS检测算法 20

4.1.1基于差分阈值的QRS检测算法 20

4.1.2基于小波变换的QRS检测算法 21

4.2基于EEMD和Hilbert变换的改进QRS检测算法 22

4.2.1ECG信号预处理 23

4.2.2基于EEMD的信号分解 23

4.2.3Hilbert变换 24

4.2.4实验结果与分析 27

4.3本章小结 29

第五章 总结与展望 30

5.1论文总结 30

5.2工作展望 30

参考文献 32

致 谢 35

攻读学士学位期间的研究成果 36

第一章 绪论

1.1研究背景与意义

心血管疾病(CVD)是有关心脏或血管的严重危害人类健康安全的一类循环系统疾病,是全球最主要的人类死亡原因之一。在许多发展中国家如中国、印度等,心血管疾病的发病率与致死率正在增加。根据2018年公布的《中国心血管疾病报告2017》估算:居民疾病死亡中的40%以上组成结构都是心血管疾病死亡,远高于肿瘤等其他重大疾病。且最近几年,农村地区心血管疾病死亡率一直高于城市地区,心血管疾病患病人数会在之后10年内快速上升。无论对于政府还是人民群众,心血管疾病的负担日渐严重,特别是农村地区的心血管疾病死亡人数会大幅增加[1]。

虽然智能手机与移动网络已较为普及,但初级卫生保健的配备仍不完善。世界各地的许多农村人口依赖由非专业志愿者组成的诊所,通过遥远城市医院的卫生保健专员来识别是否有需要进行二次医护的患者。因而,向农村诊所提供便宜的医疗仪器,例如心电图仪,成为越来越可行的方法。心脏节律性收缩与舒张的体表电活动变化是通过心电图(ECG)显示的,对诸如心律失常、心肌梗死等疾病的手术指导、药物指示有重要的临床参考价值。这些心电图仪将数字心电图传输到智能手机上存储和显示,将诊断医生的触角延伸到偏远地区。但如果没有一些质量控制手段,仅凭技术无法提供一致可用的信息。提高收集数据的质量,可以更好地辅助临床诊断。而且人们日益增长的对移动医疗的兴趣,也推动了智能手机与医疗数据结合的愿景,实现为服务不足的人群提供点对点的诊断服务。

剩余内容已隐藏,请支付后下载全文,论文总字数:41554字

您需要先支付 80元 才能查看全部内容!立即支付

该课题毕业论文、开题报告、外文翻译、程序设计、图纸设计等资料可联系客服协助查找;