基于包络提取的心音信号分段算法的研究

发布时间：2018-03-05 23:32

本文选题：心音包络　切入点：归一化香农能量　出处：《浙江大学》2015年硕士论文　论文类型：学位论文

【摘要】：心脑血管疾病是近年来发病率较高的疾病之一,不仅难以预防,而且致死率高,是人们特别是老年人的头号杀手,而心脏类疾病又是心血管疾病里比较常见的疾病。现有的检测心脏疾病的手段比较多,比如心电图、B超以及彩超等,但此类检测手段只能在心脏出现明显病变后才能检测出问题来,不利于疾病的尽早发现和治疗。另外一种检测手段——心音听诊,可以在心脏出现明显病变之前就发现相关的病变信息,但是心音听诊没有统一的客观评价标准,只能依赖于医生的主观经验。因此,本文结合心音的生理学特点,通过心音信号包络提取方法和分段方法,提取相关的医学指标参数,尝试为临床诊断提供一些客观评价依据。本文研究了不同的心音信号包络提取方法和分段方法,提取了一些医学指标参数,开发了心音分段图形用户界面,主要的工作内容如下：1.对心音信号进行了预处理。使用重采样对心音信号进行处理以减少数据量,减轻后续数据处理的压力；使用五阶巴特沃斯带通滤波器滤除心音信号中的高频和低频噪声；使用自适应小波去噪滤除部分频率与心音信号频率有重叠的噪声；使用归一化对心音信号的强度进行统一。2.提出了一种将归一化香农能量方法与希尔伯特黄变换方法相结合的心音包络提取方法。首先介绍了两种常用的包络提取方法,即归一化香农能量方法和希尔伯特黄变换方法,并通过实验对其优缺点进行了比较。为了弥补归一化香农能量方法和希尔伯特黄变换方法的一些不足,本文提出了一种将归一化香农能量方法和希尔伯特黄变换方法相结合的心音包络提取方法,在该方法中,使用镜像闭合延拓方法对端点飞翼问题进行了解决。实验表明,新方法能够得到较好的包络曲线。3.对心音信号进行了分段并提取了相关的医学指标参数。首先对单阈值分段方法及其优缺点进行了介绍,为了弥补单阈值分段方法的不足,本文对双阈值分段方法进行了研究。实验表明,双阈值分段方法较之单阈值方法具有更好的分段效果。其次介绍了一些心音相关的医学指标参数,并针对分段后的心音信号进行了医学指标参数的提取。4.开发了心音分段图形用户界面。对图形用户界面的一般设计原则和制作步骤进行了介绍,并且开发了心音分段图形用户界面。使用上述三种包络提取方法和双阈值分段方法,本文对包括309个第一心音和304个第二心音在内的30例心音信号进行了分析测试,实验结果表明,对归一化香农能量包络进行分段,第一心音和第二心音的检出率分别为84.47%和84.54%；对希尔伯特包络进行分段,第一心音和第二心音的检出率分别为74.43%和72.04%；对新包络提取方法得到的包络曲线进行分段,第一心音和第二心音的检出率分别为91.59%和90.79%。对比实验结果可知,新包络提取方法能够获得最好的分段效果。
[Abstract]:Cardio-cerebrovascular disease is one of the diseases with high incidence in recent years. It is not only difficult to prevent, but also has a high mortality rate. It is the number one killer of people, especially the elderly. And heart disease is a common disease in cardiovascular disease. There are more methods available to detect heart disease, such as electrocardiogram (ECG), B-mode ultrasound (B-ultrasound) and color Doppler ultrasound (CDUs), etc. But this kind of detection method can detect the problem only after the heart has obvious pathological changes, which is not conducive to the early detection and treatment of the disease. We can find the relevant pathological information before the obvious pathological changes in the heart, but there is no unified objective evaluation standard in the auscultation of heart sounds, which can only depend on the subjective experience of the doctors. Therefore, this paper combines the physiological characteristics of heart sounds. In this paper, the envelope extraction method and segmentation method of heart sound signal are used to extract the relevant medical index parameters and to provide some objective evaluation basis for clinical diagnosis. This paper studies different methods of heart sound signal envelope extraction and segmentation. Some medical index parameters are extracted, and the heart sound segmentation graphical user interface is developed. The main work is as follows: 1. The heart sound signal is preprocessed. The heart sound signal is processed by resampling in order to reduce the amount of data. To reduce the pressure of the subsequent data processing, the fifth order Butterworth bandpass filter is used to filter the high frequency and low frequency noise in the heart sound signal, and the adaptive wavelet denoising method is used to filter the noise with overlapping frequency of the part frequency and the heart sound signal frequency. The intensity of heart sound signal is unified by normalization. 2. A new method of extracting heart sound envelope is proposed, which combines normalized Shannon energy method with Hilbert yellow transform method. Firstly, two kinds of commonly used envelope extraction methods are introduced. That is, normalized Shannon energy method and Hilbert-Huang transform method, and their advantages and disadvantages are compared through experiments, in order to make up for some shortcomings of normalized Shannon energy method and Hilbert Huang transform method. In this paper, a method of extracting heart sound envelope by combining normalized Shannon energy method with Hilbert-Huang transform method is proposed. In this method, the problem of terminal flying wing is solved by image closed continuation method. The new method can get a better envelope curve .3.The heart sound signal is segmented and the relevant medical index parameters are extracted. Firstly, the single threshold segmentation method and its advantages and disadvantages are introduced, in order to make up for the shortcomings of the single threshold segmentation method. In this paper, the double threshold segmentation method is studied. The experimental results show that the double threshold segmentation method has better segmentation effect than the single threshold method. Secondly, some medical parameters related to heart sounds are introduced. At the same time, the medical index parameters of the segmented heart sound signal are extracted. 4. The segmentation graphical user interface of heart sound is developed. The general design principles and the manufacturing steps of the graphical user interface are introduced. Using the above three envelope extraction methods and two threshold segmentation methods, 30 heart sound signals, including 309 first heart sounds and 304 second heart sounds, were analyzed and tested. The experimental results show that the detection rates of the normalized Shannon energy envelope are 84.47% and 84.54 respectively for the first heart sound and the second heart sound, and the Hilbert envelope is segmented. The detection rates of first heart sound and second heart sound were 74.43% and 72.04 respectively. The new envelope extraction method can obtain the best segmentation effect.
【学位授予单位】：浙江大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：R54;TN912.3

【参考文献】