基于小波分析的心律失常诊断方法研究
|
摘要
心电图特别是动态心电图监测的时间长,数据量大,医生无法逐个检查心电波形。为了将医务人员从繁重、枯燥的海量心电数据处理中解放出来,避免医务人员由于环境及疲劳造成的对有效数据遗漏或误诊,心电图自动诊断技术应运而生。
在目前的心电数据分析中,只有少量的是由计算机自动分析完成的。在美国,每年约有800万例应用计算机自动分析,占10%;日本每年约有100万例,占5%;我国目前应用更少。心电诊断类产品,包括美国通用电气医疗公司、西门子医疗公司,和国内领先的迈瑞公司等,它们提供的心电监护仪器,主要功能是心电信号和常用参数的显示、数据回放、波形异常报警,以及显示ST段趋势图等。在医院采用的多是心电图检查和监护用,仅仅是用于医生观察心电图是否正常。计算机自动分析在临床分析技术中所占比例小的重要原因在于波形的测量与识别的困难和由此所造成的诊断错误。本文就是在这样的背景下产生的。
论文在广泛查阅相关文献资料的基础上,对七种常见的心律失常的心电信号的特点进行研究,提出了通过计算机诊断心律失常的一种方法。为了对心律失常进行研究,本文先对心电信号的三种主要噪声进行了分析,利用小波分解重构和阈值法对心电信号预处理,以提高信噪比,然后提取了QRS波周期等特征参数,作为分析的基础,通过检测各特征参数的变化实现心率变异性的检测和常见心律失常的诊断。
测试部分,本文先对MIT-BIH心律失常数据库中的所有记录进行了诊断,和专家给出的注释达到了较好的一致性;然后对部分实际的心律失常心电图进行了诊断,对比于医生给出的诊断结果,得到了较高的可靠性。最后,本论文提出了其后续设计和研究的一些建议。
Electrocardiogram (ECG) is a simple and practical approach for record of cardiac electrical activity. It can reflect the excitement in the heart of the dissemination process and the functional state of the heart. If the cardiac conduction system is obstacles or myocardial disease in a certain part, cardiac electrical activity can correct reflected the changes in a timely manner on the ECG, it showed as abnormal changes in the various waveform and of the evolutionary process. From 1903 ECG technology was used by Einthoven in clinical, it has been in the past 100 years, during the 100 years, the sustained development of ECG technology has made tremendous contributions to human life and health, biology, clinical medicine, it become indispensable, and most important in conventional techniques to clinical. In recent years, the cardiovascular and the cerebrovascular diseases morbidity and mortality increased year after year, to do regular checks of the heart, early detection of potential problems and has been accepted by most people, especially the needs of patients with heart disease. ECG examination as a conventional and necessary means of heart disease, is a doctor assistant diagnosis of heart disease, the accuracy of its interpretation is of great significance for the early detection of disease and improving national health standards and quality of life.
In the current ECG data analysis, only a small number of automatically completed by the computer analysis. In the United States, there are about 8 million cases a year of automatic computer analysis, accounting for 10%; In Japan, there are about 1 million cases a year, accounting for 5%; China is of less application current. ECG diagnostic products, including GE Medical in the United States, Siemens Medical Corporation, and Mairui company, which is the leading domestic company. the ECG apparatus they provided, main function is ECG parameters data playback, waveform anomalies alarm, and display such as ST segment trend. In the hospital ECG is used more by inspections and monitoring, observation is only for doctors to see if the ECG is normal. The important reason for automatic analysis based on computer technology in the clinical analysis of the small proportion is the difficulties of waveform measurement and recognition, and the resulting caused by the wrong diagnosis. The paper is built on this backdrop.
Papers were divided into six chapters, and the structure of the content is as follows:
The first chapter describes a study of computer ECG diagnosis, that is, the significance of studying automatic diagnostic ECG and its presence in the clinical significance of medical and teaching aspects. Summed up the development of through the using of computer diagnostic ECG arrhythmia and the issues, and put forth the main content of this paper. This paper introduced the simulation analysis tools: MATLAB7.5 and data sources: MIT-BIH database.
The second chapter describes the basic knowledge of the electrocardiogram. Including the formation of principle, and the characteristics of ECG, which reveals its support for the diagnostic value of arrhythmia.
The third chapter gives a pretreatment to the ECG signal which will be used for analysis. First, research on characteristics of the three major noise (baseline drift, the frequency interference, electromyography interference) of ECG. Baseline drift which was significantly lower than the signal frequency band; The frequency band of frequency interference is higher than the band of waveform we want to extract; Electromyography interference’s frequency band and the signal’s are in aliasing.. To interface these features, introducing wavelet transform theory and we use it to decompose signal. Through low-frequency signal filtering to remove baseline drift; Through high-frequency signal filtering the details in addition to frequency interference; Finally, through the threshold value method eliminates the mechanical and electrical interference.
Chapter IV, we respectively extract the waveform parameters and characteristics of the ECG. As the signal energy mainly concentrated in the third yardstick, Therefore, in the third scale wavelet decomposition, we find the R-determined position through finding extreme-right. In order to prevent misuse and missed the seizure, Design of the adaptive threshold has been done. After determining the location of R-wave, there is a appropriate fenestration before and after the R wave. We determine all other wave peak and the location of the start and finish though looking for maximum points, achieve the purpose of extracting characteristic parameters. And through MIT-BIH database validation, to good effect.
Chapter V is for HRV analysis and the diagnosis of arrhythmia. Used of the RR interval during spectral analysis of HRV, extracted the IF / HF, variance, and other important parameters and discovered abnormal heart rate. Through the ECG arrhythmia analysis of seven common characteristics, we found these arrhythmias by the computer diagnosis. Finally first diagnosis method has been applied to the data in the records, compared to the experts’Notes, Achieved better consistency; then the actual patients were diagnosis, got the comparison of results by the doctors’diagnosis, Prove the practicality of the method.
The chapter VI summarizes all this paper, briefs the job have done and the research results, and puts up some questions which should be researched continuously in the future.
在目前的心电数据分析中,只有少量的是由计算机自动分析完成的。在美国,每年约有800万例应用计算机自动分析,占10%;日本每年约有100万例,占5%;我国目前应用更少。心电诊断类产品,包括美国通用电气医疗公司、西门子医疗公司,和国内领先的迈瑞公司等,它们提供的心电监护仪器,主要功能是心电信号和常用参数的显示、数据回放、波形异常报警,以及显示ST段趋势图等。在医院采用的多是心电图检查和监护用,仅仅是用于医生观察心电图是否正常。计算机自动分析在临床分析技术中所占比例小的重要原因在于波形的测量与识别的困难和由此所造成的诊断错误。本文就是在这样的背景下产生的。
论文在广泛查阅相关文献资料的基础上,对七种常见的心律失常的心电信号的特点进行研究,提出了通过计算机诊断心律失常的一种方法。为了对心律失常进行研究,本文先对心电信号的三种主要噪声进行了分析,利用小波分解重构和阈值法对心电信号预处理,以提高信噪比,然后提取了QRS波周期等特征参数,作为分析的基础,通过检测各特征参数的变化实现心率变异性的检测和常见心律失常的诊断。
测试部分,本文先对MIT-BIH心律失常数据库中的所有记录进行了诊断,和专家给出的注释达到了较好的一致性;然后对部分实际的心律失常心电图进行了诊断,对比于医生给出的诊断结果,得到了较高的可靠性。最后,本论文提出了其后续设计和研究的一些建议。
Electrocardiogram (ECG) is a simple and practical approach for record of cardiac electrical activity. It can reflect the excitement in the heart of the dissemination process and the functional state of the heart. If the cardiac conduction system is obstacles or myocardial disease in a certain part, cardiac electrical activity can correct reflected the changes in a timely manner on the ECG, it showed as abnormal changes in the various waveform and of the evolutionary process. From 1903 ECG technology was used by Einthoven in clinical, it has been in the past 100 years, during the 100 years, the sustained development of ECG technology has made tremendous contributions to human life and health, biology, clinical medicine, it become indispensable, and most important in conventional techniques to clinical. In recent years, the cardiovascular and the cerebrovascular diseases morbidity and mortality increased year after year, to do regular checks of the heart, early detection of potential problems and has been accepted by most people, especially the needs of patients with heart disease. ECG examination as a conventional and necessary means of heart disease, is a doctor assistant diagnosis of heart disease, the accuracy of its interpretation is of great significance for the early detection of disease and improving national health standards and quality of life.
In the current ECG data analysis, only a small number of automatically completed by the computer analysis. In the United States, there are about 8 million cases a year of automatic computer analysis, accounting for 10%; In Japan, there are about 1 million cases a year, accounting for 5%; China is of less application current. ECG diagnostic products, including GE Medical in the United States, Siemens Medical Corporation, and Mairui company, which is the leading domestic company. the ECG apparatus they provided, main function is ECG parameters data playback, waveform anomalies alarm, and display such as ST segment trend. In the hospital ECG is used more by inspections and monitoring, observation is only for doctors to see if the ECG is normal. The important reason for automatic analysis based on computer technology in the clinical analysis of the small proportion is the difficulties of waveform measurement and recognition, and the resulting caused by the wrong diagnosis. The paper is built on this backdrop.
Papers were divided into six chapters, and the structure of the content is as follows:
The first chapter describes a study of computer ECG diagnosis, that is, the significance of studying automatic diagnostic ECG and its presence in the clinical significance of medical and teaching aspects. Summed up the development of through the using of computer diagnostic ECG arrhythmia and the issues, and put forth the main content of this paper. This paper introduced the simulation analysis tools: MATLAB7.5 and data sources: MIT-BIH database.
The second chapter describes the basic knowledge of the electrocardiogram. Including the formation of principle, and the characteristics of ECG, which reveals its support for the diagnostic value of arrhythmia.
The third chapter gives a pretreatment to the ECG signal which will be used for analysis. First, research on characteristics of the three major noise (baseline drift, the frequency interference, electromyography interference) of ECG. Baseline drift which was significantly lower than the signal frequency band; The frequency band of frequency interference is higher than the band of waveform we want to extract; Electromyography interference’s frequency band and the signal’s are in aliasing.. To interface these features, introducing wavelet transform theory and we use it to decompose signal. Through low-frequency signal filtering to remove baseline drift; Through high-frequency signal filtering the details in addition to frequency interference; Finally, through the threshold value method eliminates the mechanical and electrical interference.
Chapter IV, we respectively extract the waveform parameters and characteristics of the ECG. As the signal energy mainly concentrated in the third yardstick, Therefore, in the third scale wavelet decomposition, we find the R-determined position through finding extreme-right. In order to prevent misuse and missed the seizure, Design of the adaptive threshold has been done. After determining the location of R-wave, there is a appropriate fenestration before and after the R wave. We determine all other wave peak and the location of the start and finish though looking for maximum points, achieve the purpose of extracting characteristic parameters. And through MIT-BIH database validation, to good effect.
Chapter V is for HRV analysis and the diagnosis of arrhythmia. Used of the RR interval during spectral analysis of HRV, extracted the IF / HF, variance, and other important parameters and discovered abnormal heart rate. Through the ECG arrhythmia analysis of seven common characteristics, we found these arrhythmias by the computer diagnosis. Finally first diagnosis method has been applied to the data in the records, compared to the experts’Notes, Achieved better consistency; then the actual patients were diagnosis, got the comparison of results by the doctors’diagnosis, Prove the practicality of the method.
The chapter VI summarizes all this paper, briefs the job have done and the research results, and puts up some questions which should be researched continuously in the future.
引文
[1] 刘光达,赵立荣. 基于最小均方误差原理的医学 X 光影像滤波阈值选择. 光学精密工程,2001,2:47~50.
[2] 胡昌华,张军波,夏军,张伟. 基于 MATLAB 的系统分析与设计——小波分析. 西安:西安电子科技大学出版社,1999,12:1~20.
[3] 徐长发,李国宽. 实用小波方法. 武汉:华中科技大学出版社,2004,1,(2):107~119.
[4] 飞思科技产品研发中心. MATLAB6.5 辅助小波分析与应用. 北京:电子工业出版社(第二版),2003.6:6~12.
[5] 刘光达,赵立荣. 基于小波分析分析的医学 CR 影像噪声消除. 光学精密工程,2005,8(5):428~430.
[6] 梁学章,何甲兴,王新民,李强. 小波分析. 北京:国防工业出版社,2005,1:46~59.
[7] 彭玉华. 小波变换与工程应用. 北京:科学出版社,2002,2:115~132.
[8] 孙延奎. 小波分析及其应用. 北京:机械工业出版社,2005,3:221~243.
[9] 高晋占. 微弱信号检测. 北京:清华大学出版社,2004,11:3~5.
[10] 赵红怡,张常年. 数字信号处理及其 MATLAB 实现. 北京:化学工业出版社,2002,1:183~185.
[11] 梅志红,杨万铨. MATLAB 程序设计基础及其应用. 北京:清华大学出版社,2005,7:33~58.
[12] 刘群. 现代监护仪的最新应用技术. 医疗设备,2004,12:38~41.
[13] 王 保 华 , 关 晓 光 . 生 物 医 学 测 量 与 仪 器 . 上 海 : 复 旦 大 学 出 版 社 ,2003,6:292~193,301~303,140~143.
[14] 刘嘉宜,李守春. 光电技术. 雅园出版社,2000,4:83~87.
[15] 常 昌 远 , 魏 立 同 . 无 创 血 氧 饱 和 度 传 感 系 统 的 研 究 . 应 用 科 学 学 报 ,1999,12:433~438.
[16] 张小飞,徐大专,齐泽锋. 基于模极大值小波域的去噪算法研究. 数据采集与处理,2003.9,18(3):315~318.
[17] 飞思科技产品研发中心. 小波分析理论与 MATLAB7 实现. 北京:电子工业出版社,2005,11:31~39.
[18] 飞思科技产品研发中心. MATLAB7 辅助信号处理技术与应用. 北京:电子工业出版社,2005,11:29~32.
[19] [美]Emanuel Stein 编著,王瑾,姜铁民译. 心律失常快速分析. 天津科技翻译出版公司,2004,1:101~144.
[20] 王鸣和. 临床心律失常诊疗手册. 上海科学技术文献出版社,2004,5:13~23.
[21] 周伟. MATLAB 小波分析高级技术. 西安电子科技大学出版社,2006,1:105~111.
[22] L.K.Daskalov, L.I.Christov. Electrocardiogram signal preproce- ssing for automatic detection of QRS boundaries. Medical Engineering & Physics,1999,21:37-44.
[23] Tom. M,Mitchell. Machine Learning. 机械工业出版社,2003,38-48.
[24] L. Thomas. Recommendation for Standards of instrumentation and Practice in the use of ambulatory electrocardiograph Circulation. 1985,71:626.
[25]Brian Bradie. Wavelet Packet-Based Compression of Twelve Leads ECG. IEEE Transactions on Biomedical Engineering.1996,43(5):493.
[26]L. Auslander,T. Kailanth and S. Mitter. Signal Proeessing:Signal Proeessing Theory. The IMA Volumes in Mathematics and Its Applieations.1990,22.
[27]I.Daubechies. Ten Lectures on Wavelets. Philadelphia:SIAM Publish,1992:29~35.
[28]G.Beylkin,R.Coifman and I.Daubechies. Wavelets and Their Applications. Cambridge: Jones and Bartlett,MA,1992:331~348.
[29]P.Parker. Analysis of EEG transients by means of matching pursuit. Annals Biomedical Engineering.1998,23:608~611.
[30]I.Daubechies. Where do wavelets come from?-A Personal Point of view. Proc.IEEE.1996,8(45):510~513.
[31]Ahlstrom. ML, Tompkins. WJ. Digital filters for real-time ECG signal processing using microprocessors. IEEE TRANS on BME, 1985,32(9):8~15.
[32]Adriaan Ligtenberg, Murat Kunt. A robust-digital QRS-detection algorithm for arrhythmia monitoring. Computers and Biomedical Research,1983,1(6):273~276.
[33]Xue QZ, Hu Yuhen, Tompkins.WJ. Neural-network-based adaptive matched filtering for QRS detection. IEEE TRANS on BME, 1992,3(4):317~319.
[34]Cuiwei Li, Chongxun Zheng,Changfeng.Tai. Detection of ECG characteristic points using wavelet transforms. IEEE TRANS on BME,1995,42(1):21.
[35]Senhadji. Wavelet analysis of ECG signal. EEE-EMBS, 1990,12(2):811.
[36]Van Alstey JA. Removal of base-line wander and power-line interference from the ECG by an efficient FIR filter with a reduced number of taps. IEEE TRANS on BME,1985,32(12):1052.
[37]P.O.Boljesson. A daptive QRS Detection Based on Maximum, A Posteriori Estimation. IEEE Trans, Biomed. Eng, 1982,29(5):341~350.
[38]L.Sommo, O.Pahlnn, M.Nygards. A QRS Detection, A Study of Performance. IEEE Trans. Biomed. Eng, 1985,32(6):392~401.
[2] 胡昌华,张军波,夏军,张伟. 基于 MATLAB 的系统分析与设计——小波分析. 西安:西安电子科技大学出版社,1999,12:1~20.
[3] 徐长发,李国宽. 实用小波方法. 武汉:华中科技大学出版社,2004,1,(2):107~119.
[4] 飞思科技产品研发中心. MATLAB6.5 辅助小波分析与应用. 北京:电子工业出版社(第二版),2003.6:6~12.
[5] 刘光达,赵立荣. 基于小波分析分析的医学 CR 影像噪声消除. 光学精密工程,2005,8(5):428~430.
[6] 梁学章,何甲兴,王新民,李强. 小波分析. 北京:国防工业出版社,2005,1:46~59.
[7] 彭玉华. 小波变换与工程应用. 北京:科学出版社,2002,2:115~132.
[8] 孙延奎. 小波分析及其应用. 北京:机械工业出版社,2005,3:221~243.
[9] 高晋占. 微弱信号检测. 北京:清华大学出版社,2004,11:3~5.
[10] 赵红怡,张常年. 数字信号处理及其 MATLAB 实现. 北京:化学工业出版社,2002,1:183~185.
[11] 梅志红,杨万铨. MATLAB 程序设计基础及其应用. 北京:清华大学出版社,2005,7:33~58.
[12] 刘群. 现代监护仪的最新应用技术. 医疗设备,2004,12:38~41.
[13] 王 保 华 , 关 晓 光 . 生 物 医 学 测 量 与 仪 器 . 上 海 : 复 旦 大 学 出 版 社 ,2003,6:292~193,301~303,140~143.
[14] 刘嘉宜,李守春. 光电技术. 雅园出版社,2000,4:83~87.
[15] 常 昌 远 , 魏 立 同 . 无 创 血 氧 饱 和 度 传 感 系 统 的 研 究 . 应 用 科 学 学 报 ,1999,12:433~438.
[16] 张小飞,徐大专,齐泽锋. 基于模极大值小波域的去噪算法研究. 数据采集与处理,2003.9,18(3):315~318.
[17] 飞思科技产品研发中心. 小波分析理论与 MATLAB7 实现. 北京:电子工业出版社,2005,11:31~39.
[18] 飞思科技产品研发中心. MATLAB7 辅助信号处理技术与应用. 北京:电子工业出版社,2005,11:29~32.
[19] [美]Emanuel Stein 编著,王瑾,姜铁民译. 心律失常快速分析. 天津科技翻译出版公司,2004,1:101~144.
[20] 王鸣和. 临床心律失常诊疗手册. 上海科学技术文献出版社,2004,5:13~23.
[21] 周伟. MATLAB 小波分析高级技术. 西安电子科技大学出版社,2006,1:105~111.
[22] L.K.Daskalov, L.I.Christov. Electrocardiogram signal preproce- ssing for automatic detection of QRS boundaries. Medical Engineering & Physics,1999,21:37-44.
[23] Tom. M,Mitchell. Machine Learning. 机械工业出版社,2003,38-48.
[24] L. Thomas. Recommendation for Standards of instrumentation and Practice in the use of ambulatory electrocardiograph Circulation. 1985,71:626.
[25]Brian Bradie. Wavelet Packet-Based Compression of Twelve Leads ECG. IEEE Transactions on Biomedical Engineering.1996,43(5):493.
[26]L. Auslander,T. Kailanth and S. Mitter. Signal Proeessing:Signal Proeessing Theory. The IMA Volumes in Mathematics and Its Applieations.1990,22.
[27]I.Daubechies. Ten Lectures on Wavelets. Philadelphia:SIAM Publish,1992:29~35.
[28]G.Beylkin,R.Coifman and I.Daubechies. Wavelets and Their Applications. Cambridge: Jones and Bartlett,MA,1992:331~348.
[29]P.Parker. Analysis of EEG transients by means of matching pursuit. Annals Biomedical Engineering.1998,23:608~611.
[30]I.Daubechies. Where do wavelets come from?-A Personal Point of view. Proc.IEEE.1996,8(45):510~513.
[31]Ahlstrom. ML, Tompkins. WJ. Digital filters for real-time ECG signal processing using microprocessors. IEEE TRANS on BME, 1985,32(9):8~15.
[32]Adriaan Ligtenberg, Murat Kunt. A robust-digital QRS-detection algorithm for arrhythmia monitoring. Computers and Biomedical Research,1983,1(6):273~276.
[33]Xue QZ, Hu Yuhen, Tompkins.WJ. Neural-network-based adaptive matched filtering for QRS detection. IEEE TRANS on BME, 1992,3(4):317~319.
[34]Cuiwei Li, Chongxun Zheng,Changfeng.Tai. Detection of ECG characteristic points using wavelet transforms. IEEE TRANS on BME,1995,42(1):21.
[35]Senhadji. Wavelet analysis of ECG signal. EEE-EMBS, 1990,12(2):811.
[36]Van Alstey JA. Removal of base-line wander and power-line interference from the ECG by an efficient FIR filter with a reduced number of taps. IEEE TRANS on BME,1985,32(12):1052.
[37]P.O.Boljesson. A daptive QRS Detection Based on Maximum, A Posteriori Estimation. IEEE Trans, Biomed. Eng, 1982,29(5):341~350.
[38]L.Sommo, O.Pahlnn, M.Nygards. A QRS Detection, A Study of Performance. IEEE Trans. Biomed. Eng, 1985,32(6):392~401.