【作者】 谢梅兰；

【导师】 郭兴明；

【作者基本信息】 重庆大学 ， 生物医学工程， 2010， 硕士

【摘要】 心音听诊是检查心脏功能完整性的一种重要方式,可借由听诊器大概诊断出病患的心脏疾病。若要确定到某一类型的疾病上,其中一个关键的指标为心脏杂音。在青少年和婴幼儿人群的心脏检查中,心脏杂音是最常见的体征,如不能做出正确的临床诊断和分析,将会给病人及家属造成很大的精神压力。心脏杂音可通过其强度,持续时间和音质等特征来分级。传统的杂音分级方法源于1933年Freeman和Levine提出的I-VI级杂音分级系统。该方法由内科医生依据自身的经验做出直接、主观的判断,其判断结果会因为检查者自身的情况而改变。同时,医生常常采用“吹风样杂音”、“隆隆样杂音”等术语来描述心脏杂音的特点。显然这些术语只提供了主观的依据,从中难以得到关于心脏疾病准确的生理、病理特征。本论文通过介绍心脏杂音相关生理基础(如心脏杂音产生机制、特征量等)及其在心脏疾病诊断中的应用,在此基础上进一步对杂音信号进行预处理,提取特征参数作为杂音分级量化指标,最后从心脏杂音不同频率成分能量分布和不同心音成分能量分数两方面对其能量状态进行分析。心音信号预处理主要包括小波阈值消噪及Welch功率谱估计。首先,介绍噪声在小波分解下的特性及小波阈值降噪的方法和步骤,通过小波函数的性质与心音信号特点结合分析,基于重构因子方法选取出适合的小波函数进行小波阈值消噪。试验结果证明,coif3小波函数对心音的降噪效果良好。利用Welch功率谱估计方法对心音进行谱估计,得到不同成分的频域特征,作为心音分段和定位的基础。由于小波变换具有良好的多分辨率分析特性,利用MALLAT算法进行心脏杂音隔离,在此基础上通过希尔伯特变换获得心音包络,差分方法提取包络峰值,设置一个合理的阈值抛弃多余的峰值点,最后在包络峰值点周围的1000个采样点内寻找第一心音(S1),第二心音(S2)边界点,从而确定S1,S2的持续时间。在此基础上,进一步从小波变换的不同子带滤波信号中提取三个连续心动周期内杂音和S1,S2与基线分别围成的面积、持续时限,最大幅值,能量等参数,计算三个周期内杂音与S1,S2(作参考)比值指标的均值,得到三种心脏杂音分级量化的方法(杂音面积比值法,杂音强度-时限加权法和杂音能量比值法)。结合传统心脏杂音分级结果,对比三种量化指标的优异,为心脏杂音从主观分级到客观分级提供一种主观和客观相结合的过渡研究。为了解不同类型心杂音信号不同频率成分的能量分布情况,本文利用小波包分解分析正常心音和各类型心杂音信号能量特征。由于泵血的力来源于心脏收缩,S1幅值的变化与左室压力上升的最大速率密切相关,S1能量能在一定程度上反映心脏能量的消耗,对应于有用功。从效率的观点看,返流性心脏杂音对应于无用功。分别计算三个连续心动周期的S1,杂音成分相对于能量总和的占有分数,即第一心音能量分数(S1EF)和杂音能量分数(MEF),从能量效率角度评估正常人群和返流性心杂音人群的心脏能量状态。试验结果表明,在正常组和返流性心脏杂音组间,第一心音能量分数存在显著性差异。更多还原

【Abstract】 Heart sound auscultation is an important way to check the integrity of cardiac function, and the patients with heart diseases can be probably diagnosed by means of the stethoscope. To determine a certain type of disease, one of the crucial indicators is heart murmur. Heart murmur is the most common signs in the heart examination of young people and infants, if the examiner can not make a correct clinical diagnosis and analysis, it will be caused a great stress to patients and their families. Heart murmurs can be graded as to intensity, duration, and quality. Customarily, intensity or loudness is graded from one to six, based upon the original recommendations of Freeman and Levine in 1933. The method is based on the physician experience to make direct, subjective judgments, and the results will be changed with the examiner situation. Meanwhile, physicians often use "blowing", "rumbling" and other terms to describe the characteristics of the heart murmur. Obviously, these terms provides only subjective reference, while it’s hard to obtain accurate physiological and pathological characteristics from heart diseases.This paper first introduces the physiological basis of heart murmur (such as heart murmur generation mechanism, features, etc.) and the applications in the diagnosis of heart disease. Based on murmurs feature, a further signal preprocessing will be needed. Then characteristic parameters are extracted as the quantitative indicators of heart murmurs grading. Finally, cardiac energy was analysis through the energy distribution of different frequency components in heart sound and the different components of their energy fractions.The pretreatment includes wavelet threshold de-noising and welch power spectrum estimation. First, the paper introduces the characteristics of noises in the wavelet decomposition, as well as the methods and procedures of wavelet threshold de-noising. Combined with the nature of wavelet functions and heart sounds characteristics, we selected the suitable wavelet function based on the reconstructed factor. The result showed that coif3 wavelet had a good de-noising for heart sounds. Moreover, the spectrum of heart sound is obtained by using welch power spectrum estimation method. It can display frequency domain characteristics in different components and pay a basis role for heart sound segmentation and positioning. We used MALLAT algorithm to isolate murmurs from heart sounds with the good multi-resolution analysis features of wavelet transform. Heart sound envelope was extracted by Hilbert transform. Differential method was used to obtain the peak values and extra peaks were discarded according to threshold. Finally, we looked for the first heart sound, the second heart sound boundaries in the peak around 1000 samples and determined duration of the first heart sound, the second heart sound. The parameters of murmur, the first heart sound and the second heart sound in three consecutive cardiac cycles were extracted,such as the area enclosed with baseline, duration, maximum amplitude, etc. By calculating the mean ratio of murmur and the first heart sound, the second heart sound (for reference), we explored three quantitative methods (murmur area ratio, weighted intensity and duration ratio, murmur energy ratio) for heart murmur grading. The results of these methods were compared with traditional murmur grading, and it provided a combination of subjective and objective study for heart murmur grading.To understand the energy distribution of different frequency components in murmurs, this paper selected wavelet packet decomposition to analysis normal heart sounds and different types of heart murmur signals. Since the force pumping the blood is derived from cardiac contraction and changes in the amplitude of the first heart sound are closely related with the maximum rate of rise of left ventricular pressure, the energy of the first heart sound can reflect cardiac energy expenditure at a certain extent, corresponding to useful work. From the viewpoint of efficiency, the energy of cardiac murmur corresponds to useless work. The first heart sound energy fraction (S1EF) and murmur energy fraction (MEF) were calculated for three consecutive cardiac cycles, the cardiac energy state was assessed in different groups by energy efficiency. The results showed that the first heart sound energy fraction were significantly different in the normal group and the heart murmur group.更多还原