WO2024007152A1 - 一种基于心电心音的儿童心血管疾病诊断方法 - Google Patents
一种基于心电心音的儿童心血管疾病诊断方法 Download PDFInfo
- Publication number
- WO2024007152A1 WO2024007152A1 PCT/CN2022/103931 CN2022103931W WO2024007152A1 WO 2024007152 A1 WO2024007152 A1 WO 2024007152A1 CN 2022103931 W CN2022103931 W CN 2022103931W WO 2024007152 A1 WO2024007152 A1 WO 2024007152A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- heart sound
- heart
- signal
- cardiovascular diseases
- features
- Prior art date
Links
- 238000000034 method Methods 0.000 title claims abstract description 55
- 208000024172 Cardiovascular disease Diseases 0.000 title claims abstract description 43
- 238000012545 processing Methods 0.000 claims abstract description 7
- 239000000284 extract Substances 0.000 claims abstract description 5
- 230000005236 sound signal Effects 0.000 claims description 48
- 238000001228 spectrum Methods 0.000 claims description 40
- 230000003205 diastolic effect Effects 0.000 claims description 14
- 210000002569 neuron Anatomy 0.000 claims description 14
- BLRBOMBBUUGKFU-SREVYHEPSA-N (z)-4-[[4-(4-chlorophenyl)-5-(2-methoxy-2-oxoethyl)-1,3-thiazol-2-yl]amino]-4-oxobut-2-enoic acid Chemical compound S1C(NC(=O)\C=C/C(O)=O)=NC(C=2C=CC(Cl)=CC=2)=C1CC(=O)OC BLRBOMBBUUGKFU-SREVYHEPSA-N 0.000 claims description 13
- 238000004458 analytical method Methods 0.000 claims description 12
- 230000000747 cardiac effect Effects 0.000 claims description 12
- 238000001914 filtration Methods 0.000 claims description 8
- 230000009467 reduction Effects 0.000 claims description 6
- 230000008569 process Effects 0.000 claims description 4
- 238000007781 pre-processing Methods 0.000 claims description 3
- 238000003745 diagnosis Methods 0.000 abstract description 5
- 230000006870 function Effects 0.000 description 4
- 238000003062 neural network model Methods 0.000 description 4
- 208000002330 Congenital Heart Defects Diseases 0.000 description 3
- 208000001910 Ventricular Heart Septal Defects Diseases 0.000 description 3
- 208000028831 congenital heart disease Diseases 0.000 description 3
- 201000003130 ventricular septal defect Diseases 0.000 description 3
- 238000002555 auscultation Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000011156 evaluation Methods 0.000 description 2
- 230000004217 heart function Effects 0.000 description 2
- 210000004115 mitral valve Anatomy 0.000 description 2
- 230000002159 abnormal effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000002457 bidirectional effect Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 238000002405 diagnostic procedure Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000003862 health status Effects 0.000 description 1
- 208000019622 heart disease Diseases 0.000 description 1
- 210000005240 left ventricle Anatomy 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000011946 reduction process Methods 0.000 description 1
- 239000000523 sample Substances 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 230000006403 short-term memory Effects 0.000 description 1
- 230000003595 spectral effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000001356 surgical procedure Methods 0.000 description 1
- 238000012549 training Methods 0.000 description 1
- 208000019553 vascular disease Diseases 0.000 description 1
- 230000002861 ventricular Effects 0.000 description 1
Images
Classifications
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/01—Protocols
- H04L67/10—Protocols in which an application is distributed across nodes in the network
- H04L67/1095—Replication or mirroring of data, e.g. scheduling or transport for data synchronisation between network nodes
Definitions
- the present invention relates to the technical field of cardiovascular disease diagnosis, and in particular to a method for diagnosing cardiovascular disease in children based on electrocardiogram and heart sounds.
- VSD ventricular Septal Defect
- the present invention provides a method for diagnosing cardiovascular diseases in children based on electrocardiogram and heart sounds; the method includes the following steps:
- S2 Process the electrocardiogram signal and heart sound signal to obtain multi-dimensional heart sound characteristics
- S3 Input the multi-dimensional heart sound features into the diagnostic model to obtain a diagnostic result, where the diagnostic result is the type of cardiovascular disease.
- the ECG signal and the heart sound signal are processed to obtain the ECG and heart sound characteristics.
- Methods include:
- the multidimensional heart sound features include time domain heart sound features, frequency domain heart sound features, and energy domain heart sound features.
- the method for preprocessing the ECG signal and the heart sound signal includes:
- S212 Perform filtering and noise reduction processing on the sampled ECG signals and heart sound signals
- S213 Normalize the ECG signal and heart sound signal after filtering and noise reduction.
- the method for obtaining initial heart sound features based on the preprocessed ECG signal and heart sound signal includes:
- S222 According to the Q wave of the QRS wave, identify the duration and amplitude information of the first heart sound from the preprocessed heart sound signal, and according to the T wave of the QRS wave, identify the second heart sound signal from the preprocessed heart sound signal. Duration and amplitude information of heart sounds;
- S223 According to the duration and amplitude information of the first heart sound and the second heart sound, obtain the duration and amplitude information of the systolic phase and the duration and amplitude information of the diastolic phase;
- S224 Use the duration and amplitude information of the first heart sound, the second heart sound, systole and diastole as initial heart sound features.
- the initial heart sound characteristics are analyzed through a time domain analysis method.
- the method of obtaining the time domain heart sound characteristics includes: respectively calculating the first heart sound, systole, second heart sound and diastole. Amplitude ratio; calculate the time difference between adjacent first heart sound, systole, second heart sound and diastole respectively; calculate EMAT%; calculate the average of each amplitude ratio, each time difference and EMAT% in the entire cardiac cycle, and get 11 time domain heart sound features.
- the initial heart sound features are analyzed through a frequency domain analysis method, and the method for obtaining the time domain heart sound features includes: performing Fourier transform on the initial heart sound features to obtain a corresponding spectrum sequence, wherein:
- the spectrum sequence includes a first heart rate spectrum sequence, a second heart rate spectrum sequence, a systolic spectrum sequence, and a diastolic spectrum sequence; the first heart rate spectrum sequence, the second heart rate spectrum sequence, the systolic spectrum sequence, and the diastolic spectrum are extracted respectively.
- the frequencies in the sequence are multiple values of specified frequencies, and the average value of each value in all cardiac cycles is calculated and a new spectrum sequence is formed; 4 frequency domain heart sound characteristics are obtained.
- the plurality of designated frequencies corresponding to the first heartbeat spectrum sequence, the second heartbeat spectrum sequence, the systolic phase, and the diastolic phase are 60HZ, 70HZ, 80HZ, 90HZ, 100HZ, 110HZ, 120HZ, 130HZ, 80HZ, and 150HZ.
- the initial heart sound characteristics are analyzed through a frequency domain analysis method, and the method for obtaining the time domain heart sound characteristics includes: separately calculating the first heart sound, systolic period, second heart sound and diastolic period. Energy ratio, and find the average value of each energy ratio in all cardiac cycles to obtain 6 energy domain heart sound characteristics.
- the diagnostic model includes an input layer, a hidden layer, an output layer and a softmax function, wherein the input layer consists of 21 neurons, and the 21 neurons correspond to multi-dimensional heart sound features; the hidden The layers have 7 and 8 neurons respectively; the output layer consists of 8 neurons, corresponding to the types of cardiovascular diseases; finally, a softmax function is used to create a multi-classification problem.
- embodiments of the present invention provide a method for diagnosing cardiovascular diseases in children based on electrocardiogram and heart sounds, which combines heart sound signals and electrocardiogram signals and fully considers the manifestation of cardiovascular diseases in heart sound signals.
- Multi-dimensional heart sound features including time domain, frequency domain and energy domain heart sound features are extracted, and accurate prediction of cardiovascular diseases is achieved by building a diagnostic model and multi-dimensional heart sound features.
- the present invention also provides a method for extracting multi-dimensional heart sound features, which provides basic support for the diagnostic model to accurately predict cardiovascular diseases.
- the invention has the advantages of practicality, high efficiency and the like.
- Figure 1 is a flow chart of a method for diagnosing cardiovascular diseases in children based on electrocardiogram and heart sounds provided by an embodiment of the present invention
- Figure 2 is a construction diagram of a diagnostic model provided by an embodiment of the present invention.
- an embodiment of the present invention provides a method for diagnosing cardiovascular disease in children based on electrocardiogram and heart sounds; the method includes the following steps:
- Step 1 Obtain the ECG signal and heart sound signal of the child to be diagnosed
- the method of obtaining the ECG signal and heart sound signal of the child to be diagnosed is to place the twelve leads of the ECG according to the conventional ECG operation method, and on this basis, simultaneously place the high-precision heart sound probe and simultaneously record the characteristics of the heart sound. Obtain the ECG signal and heart sound signal of the child to be diagnosed.
- Step 2 Process the electrocardiogram signal and heart sound signal to obtain multi-dimensional heart sound characteristics
- the method of processing the ECG signal and the heart sound signal to obtain the ECG and heart sound characteristics includes:
- Preprocess the ECG signal and heart sound signal includes: intercepting the ECG signal and heart sound signal; performing sampled ECG signal and heart sound signal. Filtering and noise reduction processing; normalizing the ECG signal and heart sound signal after filtering and noise reduction;
- the filtering and noise reduction process uses an infinite impulse corresponding digital band group filter for filtering. This filter is an existing technology, and its working process will not be described in detail here.
- the method is specifically to identify the QRS wave of the preprocessed ECG signal; based on the Q wave of the QRS wave, identify the preprocessed heart sound signal
- the duration and amplitude information of the first heart sound is obtained, and the duration and amplitude information of the second heart sound are identified from the preprocessed heart sound signal according to the T wave of the QRS wave; according to the first heart sound and the second heart sound
- the duration and amplitude information of the heart sound is obtained, and the duration and amplitude information of the systolic phase and the duration and amplitude information of the diastolic phase are obtained; the duration of the first heart sound, the second heart sound, the systolic phase and the diastolic phase are and amplitude information as initial heart sound features;
- the multidimensional heart sound features include time domain heart sound features, frequency domain heart sound features, and energy domain heart sound features.
- the initial heart sound characteristics are analyzed through the time domain analysis method, and the method of obtaining the time domain heart sound characteristics includes: separately calculating the amplitudes between the first heart sound, systole, second heart sound and diastole. Ratio; calculate the time difference between adjacent first heart sound, systole, second heart sound and diastole respectively; calculate EMAT% of each cardiac cycle respectively; calculate each amplitude ratio, each time difference and each EMAT%; get 11 time domain heart sound characteristics.
- each amplitude ratio, each time difference and each EMAT% mentioned above in each cardiac cycle are calculated to construct 6 amplitude ratio sequences, 4 time difference sequences, and 1 EMAT% sequence, and 11 time domain heart sound features are obtained. .
- EMAT is the time from the onset of the QRS wave in the electrocardiogram to the onset of the first heart sound (mitral valve closure), including the electro-mechanical delay time and the left ventricular systolic period before mitral valve closure. It is the presystolic period of the left ventricle.
- EMAT% refers to the proportion of EMAT in the RR interval. Studies have shown that EMAT% has a significant impact on cardiovascular disease. Therefore, the embodiment of the present invention uses the EMAT% sequence as a time-domain heart sound feature.
- the sequence length is a multiple of the preset sequence length
- the sequence will be segmented according to the multiple, and the average value of each value in each segment will be calculated as the new sequence.
- the sequence length is not the preset sequence length
- Multiples of the sequence delete the first value of the sequence until the sequence length is a multiple of the preset sequence length.
- the sequence length provided by the embodiment of the present invention is 10, and the cardiac cycle intercepted in the previous step is 12, then the length of the constructed time difference sequence is 11.
- the first value of the time length sequence is deleted to obtain a new time difference sequence.
- the new time difference sequence is 1 times the length of the preset sequence, then use 1 as the segmentation parameter to divide the time difference sequence into 10 segments, calculate the average of the 10 segments, and obtain a time difference sequence with a length of 10.
- the method of analyzing the initial heart sound features through the frequency domain analysis method to obtain the time domain heart sound features includes: performing Fourier transform on the initial heart sound features to obtain the corresponding spectrum sequence, wherein the spectrum
- the sequence includes the first heart rate spectrum sequence, the second heart rate spectrum sequence, the systolic spectrum sequence and the diastolic spectrum sequence; respectively extract the first heart rate spectrum sequence, the second heart rate spectrum sequence, the systolic spectrum sequence and the diastolic spectrum sequence.
- the frequency is a number of values at specified frequencies. The average value of each value in all cardiac cycles is calculated and a new spectrum sequence is formed; 4 frequency domain heart sound features are obtained.
- the spectrum distribution of the first heart sound and the second heart sound is 50-100HZ, if it exceeds 100HZ, it is determined that there is a murmur, and the preset sequence length is 10. Therefore, the five designated frequencies selected in the embodiment of the present invention
- the normal group and the abnormal group of 5 specified frequencies construct spectral heart sound characteristics.
- the multiple specified frequencies corresponding to the first heart sound spectrum sequence, the second heart sound spectrum sequence, the systolic period, and the diastolic period are 60HZ, 70HZ, 80HZ, and 90HZ. , 100HZ, 110HZ, 120HZ, 130HZ, 80HZ, 150HZ, and thus obtain 4 frequency domain heart sound features with a length of 10.
- the method of analyzing the initial heart sound characteristics through the frequency domain analysis method and obtaining the time domain heart sound characteristics includes: separately calculating the energy ratio between the first heart sound, systole, second heart sound and diastole. , and calculate the average value of each energy ratio in all cardiac cycles to obtain 6 energy domain heart sound characteristics.
- the method for processing the sequence length is the same as the above-mentioned method for obtaining the sequence length for transmitting the time domain heart sound feature, and will not be described again here.
- Step 3 Input the multi-dimensional heart sound features into the diagnostic model to obtain a diagnostic result, where the diagnostic result is the type of cardiovascular disease.
- the diagnostic model includes an input layer, a hidden layer, an output layer and a softmax function, wherein the input layer is composed of 21 neurons, and the 21 neurons correspond to Multi-dimensional heart sound features; the hidden layer has 7 and 8 neurons respectively, and the hidden layer is a 2-layer bidirectional long short-term memory network model; the output layer is composed of 8 neurons, corresponding to the type of cardiovascular disease; Finally, a softmax function is used, which is created to solve multi-classification problems.
- the output result is the probability of multiple cardiovascular diseases. The probability is a value between 0-1, and the sum of the obtained probabilities is 1.
- the number of neurons in the output layer can be 8, 50, or 100, and can be set according to existing common cardiovascular diseases. If the output layer neurons are added, diagnostic model training needs to be added. time samples to ensure the accuracy of diagnosis by the diagnostic model.
- the diagnostic model may be a fully connected neural network model, with two fully connected layers as hidden layers of the fully connected neural network model, and an S-shaped growth curve layer as the output layer of the fully connected neural network model; when the The diagnostic model is a fully connected neural network model, and its output results are the probabilities of multiple types of cardiovascular diseases.
- the probabilities are 0-1 values, and the sum of the output probabilities is not necessarily 1.
- embodiments of the present invention provide a method for diagnosing cardiovascular diseases in children based on ECG heart sounds, which combines heart sound signals and ECG signals, fully considers the manifestation of cardiovascular diseases in heart sound signals, and extracts time domain , multi-dimensional heart sound features of frequency domain and energy domain heart sound features, and by building diagnostic models and multi-dimensional heart sound features, accurate prediction of cardiovascular diseases is achieved.
- the present invention also provides a method for extracting multi-dimensional heart sound features, which provides basic support for the diagnostic model to accurately predict cardiovascular diseases.
- the invention has the advantages of practicality, high efficiency and the like.
Landscapes
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Engineering & Computer Science (AREA)
- Medical Informatics (AREA)
- Surgery (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Heart & Thoracic Surgery (AREA)
- Physics & Mathematics (AREA)
- Molecular Biology (AREA)
- Pathology (AREA)
- Animal Behavior & Ethology (AREA)
- General Health & Medical Sciences (AREA)
- Public Health (AREA)
- Veterinary Medicine (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Measuring Pulse, Heart Rate, Blood Pressure Or Blood Flow (AREA)
Abstract
一种基于心电心音的儿童心血管疾病诊断方法,涉及心血管疾病诊断技术领域,方法包括获取待诊断患儿的心电信号和心音信号(S1);对心电信号和心音信号进行处理,得到多维心音特征(S2);将多维心音特征输入诊断模型,得到诊断结果(S3)。基于心电心音的儿童心血管疾病诊断方法结合了心音信号和心电信号,充分考虑了心血管疾病在心音信号上的体现,提取了包括时域、频域和能量域心音特征的多维心音特征,通过构建诊断模型和多维心音特征,实现了心血管疾病的精准预测。另外,还提供了一种多维心音特征的提取方法,为诊断模型进行心血管疾病精准预测提供了基础支持,具有实用高效等优点。
Description
本发明涉及心血管疾病诊断技术领域,具体涉及一种基于心电心音的儿童心血管疾病诊断方法。
据统计,目前1000名新生儿中,大概有8人患有先天性心脏病,其中心室间隔缺损(Ventricular Septal Defect,VSD)是最为常见的先天性心脏病。根据心室间隔缺损患者的情况,有时必须及早进行手术,因此先天性心脏病这类心血管疾病的精确诊断,对儿童具有重大意义。
在心血管疾病的诊断方法中,听诊古时候就有,即根据心音判别心脏的健康状况,但是听诊对于医生来说,需要多年的经验积累。近年来随着计算机技术的不断发展,可以利用计算机从心音信号中抽取与心脏病或心脏功能相关的特征参数。因此,采用计算机进行心血管疾病的判别和心脏功能的评价成为可能。
目前,大多从心音信号中提取特征波形的峰值的时间幅度、时间间隔作为评价指标,进行血管疾病的判断,并没有考虑到心电信号和心音信号的结合,且评价的维度也局限于时域,没有考虑到频域、能量域等维度,因此,本申请提出了一种基于心电心音的儿童心血管疾病诊断方法。
发明内容
针对现有技术中的缺陷,本发明提供了一种基于心电心音的儿童心血管疾病诊断方法;所述方法包括以下步骤:
S1:获取待诊断患儿的心电信号和心音信号;
S2:对所述心电信号和心音信号进行处理,得到多维心音特征;
S3:将所述多维心音特征输入诊断模型,得到诊断结果,其中,所述诊断结果为心血管疾病类型。
优选地,S2中,对心电信号和心音信号进行处理,得到心电心音特征的
方法包括:
S21:对心电信号和心音信号进行预处理;
S22:根据预处理后的心电信号和心音信号,获取初始心音特征;
S23:通过时域、频域、能量域分析方法对所述初始心音特征进行分析,得到多维心音特征,所述多维心音特征包括时域心音特征、频域心音特征和能量域心音特征。
优选地,S21中,对心电信号和心音信号进行预处理的方法包括:
S211:对心电信号和心音信号进行截取;
S212:对采样后的心电信号和心音信号进行滤波降噪处理;
S213:对滤波降噪后的心电信号和心音信号进行归一化处理。
优选地,S22中,根据预处理后的心电信号和心音信号,获取初始心音特征的方法包括:
S221:识别预处理后的心电信号的QRS波;
S222:根据QRS波的Q波,从预处理后的心音信号中识别出第一心音的持续和幅值信息,根据所述QRS波的T波,从预处理后的心音信号中识别出第二心音的持续时间和幅值信息;
S223:根据第一心音和第二心音的持续时间和幅值信息,获取收缩期的持 续时间和幅值信息、舒张期的持续时间和幅值信息;
S224:将所述第一心音、第二心音、收缩期和舒张期的持续时间和幅值信息作为初始心音特征。
优选地,S23中,通过时域分析方法对所述初始心音特征进行分析,得到时域心音特征的方法包括:分别计算第一心音、收缩期、第二心音和舒张期两两之间的幅值比;分别计算相邻第一心音、收缩期、第二心音和舒张期之间的时间差;计算EMAT%;计算各个幅值比、各个时间差、EMAT%在整个心动周期的均值,得到11个时域心音特征。
优选地,S23中,通过频域分析方法对所述初始心音特征进行分析,得到时域心音特征的方法包括:对所述初始心音特征进行傅里叶变换,得到对应的频谱序列,其中,所述频谱序列包括第一心音频谱序列、第二心音频谱序列、收缩期频谱序列和舒张期频谱序列;分别提取第一心音频谱序列、第二心音频谱序列、收缩期频谱序列和舒张期频谱序列中频率为多个指定频率的数值,求取各个数值在所有心动周期的均值,并形成新的频谱序列;得到4个频域心音特征。
优选地,所述第一心音频谱序列、第二心音频谱序列、收缩期、舒张期对应的多个指定频率为60HZ、70HZ、80HZ、90HZ、100HZ、110HZ、120HZ、130HZ、80HZ、150HZ。
优选地,S23中,通过频域分析方法对所述初始心音特征进行分析,得到时域心音特征的方法包括:分别计算第一心音、收缩期、第二心音和舒张期两两之间的能量比,并求各个能量比在所有心动周期的均值,得到6个能量域心音特征。
优选地,S3中,所述诊断模型包括输入层、隐藏层、输出层和softmax函数,其中,所述输入层由21个神经元组成,所述21个神经元对应多维心音 特征;所述隐藏层分别有7和8个神经元;所述输出层由个8神经元组成,对应心血管疾病的类型;最后使用一个softmax函数,用于解决多分类问题而创建。
本发明的有益效果体现在:本发明实施例提供了一种基于心电心音的儿童心血管疾病诊断方法,结合了心音信号和心电信号,充分考虑了心血管疾病在心音信号上的体现,提取了包括时域、频域和能量域心音特征的多维心音特征,通过构建诊断模型和多维心音特征,实现了心血管疾病的精准预测。另外,本发明还提供了一种多维心音特征的提取方法,为诊断模型进行心血管疾病精准预测提供了基础支持。本发明具体实用高效等优点。
为了更清楚地说明本发明具体实施方式或现有技术中的技术方案,下面将对具体实施方式或现有技术描述中所需要使用的附图作简单地介绍。在所有附图中,类似的元件或部分一般由类似的附图标记标识。附图中,各元件或部分并不一定按照实际的比例绘制。
图1为本发明实施例所提供的一种基于心电心音的儿童心血管疾病诊断方法的流程图;
图2为本发明实施例所提供的诊断模型的构建图。
下面将结合附图对本发明技术方案的实施例进行详细的描述。以下实施例仅用于更加清楚地说明本发明的技术方案,因此只作为示例,而不能以此来限制本发明的保护范围。
需要注意的是,除非另有说明,本申请使用的技术术语或者科学术语应当为本发明所属领域技术人员所理解的通常意义。
如图1所示,本发明实施例提供了一种基于心电心音的儿童心血管疾病诊 断方法;所述方法包括以下步骤:
步骤一:获取待诊断患儿的心电信号和心音信号;
需要说明的,获取待诊断患儿的心电信号和心音信号的方法为按照常规心电图操作方法将心电图十二导联放好,在此基础上同步放置高精心音探头,同步记录心音的特征,得到待诊断患儿的心电信号和心音信号。
步骤二:对所述心电信号和心音信号进行处理,得到多维心音特征;
在本发明实施例中,对心电信号和心音信号进行处理,得到心电心音特征的方法包括:
A:对心电信号和心音信号进行预处理;其中,对心电信号和心音信号进行预处理的方法包括:对心电信号和心音信号进行截取;对采样后的心电信号和心音信号进行滤波降噪处理;对滤波降噪后的心电信号和心音信号进行归一化处理;
具体的,对心电信号和心音信号进行截取时,需要保证被截取的心电信号和心音信号至少包括11个心动周期,以满足诊断模型的输入序列的长度需求。滤波降噪处理采用无限冲激相应数字带组滤波器进行滤波,该滤波器为现有技术,其工作过程在此不做赘述。
B:根据预处理后的心电信号和心音信号,获取初始心音特征;方法具体为识别预处理后的心电信号的QRS波;根据QRS波的Q波,从预处理后的心音信号中识别出第一心音的持续和幅值信息,根据所述QRS波的T波,从预处理后的心音信号中识别出第二心音的持续时间和幅值信息;根据第一心音和第二心音的持续时间和幅值信息,获取收缩期的持续时间和幅值信息、舒张期的持续时间和幅值信息;将所述第一心音、第二心音、收缩期和舒张期的持续时间和幅值信息作为初始心音特征;
C:通过时域、频域、能量域分析方法对所述初始心音特征进行分析,得到 多维心音特征,所述多维心音特征包括时域心音特征、频域心音特征和能量域心音特征。
需要说明的,通过时域分析方法对所述初始心音特征进行分析,得到时域心音特征的方法包括:分别计算第一心音、收缩期、第二心音和舒张期两两之间的幅值比;分别计算相邻第一心音、收缩期、第二心音和舒张期之间的时间差;分别计算每个心动周期的EMAT%;计算各个幅值比、各个时间差、各个EMAT%;得到11个时域心音特征。
具体的,计算每个心动周期中上述所述的各个幅值比和各个时间差和各个EMAT%构建6个幅值比序列、4个时间差序列、1个EMAT%序列,得到11个时域心音特征。
其中,EMAT是从心电图中QRS波起始至第一心音开始(二尖瓣关闭)的时间,包括电-机械延迟时间和二尖瓣关闭前的左室收缩期,是左室收缩前期的一部分,EMAT%是指EMAT在RR间期所占的比例,经研究表明,EMAT%对心血管疾病的具有显著影响,因此,本发明实施例将EMAT%序列作为一个时域心音特征。
为保证序列长度的一致性,若序列长度是预设序列长度的倍数,则根据倍数对序列进行分段,求取各个分段中各个数值的平均值作为新的序列,若序列长度不是预设序列的倍数,则从序列的第一数值进行删除,直到序列长度为预设序列长度的倍数。具体的,本发明实施例所提供的序列长度为10,之前步骤截取的心动周期为12,则构建的时间差序列长度为11,此时删除时间长度序列的第一个数值,得到新的时间差序列,新的时间差序列为预设序列长度的1倍,则将1作为分段参数将时间差序列分为10段,求取10段的平均值,得到长度为10的时间差序列。
需要说明的,通过频域分析方法对所述初始心音特征进行分析,得到时域 心音特征的方法包括:对所述初始心音特征进行傅里叶变换,得到对应的频谱序列,其中,所述频谱序列包括第一心音频谱序列、第二心音频谱序列、收缩期频谱序列和舒张期频谱序列;分别提取第一心音频谱序列、第二心音频谱序列、收缩期频谱序列和舒张期频谱序列中频率为多个指定频率的数值,求取各个数值在所有心动周期的均值,并形成新的频谱序列;得到4个频域心音特征。
由于第一心音和第二心音的频谱分布是50~100HZ,若超过100HZ就判定为存在杂音,并且预设的序列长度为10,因此,在本发明实施例中选定的5个指定频率的正常组和5个指定频率的异常组构建频谱心音特征,所述第一心音频谱序列、第二心音频谱序列、收缩期、舒张期对应的多个指定频率为60HZ、70HZ、80HZ、90HZ、100HZ、110HZ、120HZ、130HZ、80HZ、150HZ,并从而得到4个长度为10的频域心音特征。
需要说明的,通过频域分析方法对所述初始心音特征进行分析,得到时域心音特征的方法包括:分别计算第一心音、收缩期、第二心音和舒张期两两之间的能量比,并求各个能量比在所有心动周期的均值,得到6个能量域心音特征。为保证能量域心音特征的长度满足要求,对序列长度进行处理的方法与上述获取用于输对时域心音特征序列长度进行处理的方法相同,在此不做赘述。
步骤三:将所述多维心音特征输入诊断模型,得到诊断结果,其中,所述诊断结果为心血管疾病类型。
如图2所示,在本发明实施例中,所述诊断模型包括输入层、隐藏层、输出层和softmax函数,其中,所述输入层由21个神经元组成,所述21个神经元对应多维心音特征;所述隐藏层分别有7和8个神经元,所述隐藏层为2层双向长短期记忆网络模型;所述输出层由个8个神经元组成,对应心血管疾病的类型;最后使用一个softmax函数,用于解决多分类问题而创建,其输出结果为多个心血管疾病的概率,其概率为0-1之间的数值,且得到的概率之和为1。
在其他实施例中,所述输出层的神经元可以是8个也可以是50个或者100个,可以根据现有的常见心血管疾病进行设置,若增加输出层神经元就需要增加诊断模型训练时的样本,以保证诊断模型诊断的精确性。
在其他实施例中,所述诊断模型可以是全连接神经网络模型,2层全连接层作为全连接神经网络模型的隐藏层,S型生长曲线层作为全连接神经网络模型的输出层;当所述诊断模型为全连接神经网络模型是,其输出结果为多个类型的心血管疾病的概率,其概率为0-1的数值,且输出的概率的和不一定为1。
综上,本发明实施例提供了一种基于心电心音的儿童心血管疾病诊断方法,结合了心音信号和心电信号,充分考虑了心血管疾病在心音信号上的体现,提取了包括时域、频域和能量域心音特征的多维心音特征,通过构建诊断模型和多维心音特征,实现了心血管疾病的精准预测。另外,本发明还提供了一种多维心音特征的提取方法,为诊断模型进行心血管疾病精准预测提供了基础支持。本发明具体实用高效等优点。
最后应说明的是:以上各实施例仅用以说明本发明的技术方案,而非对其限制;尽管参照前述各实施例对本发明进行了详细的说明,本领域的普通技术人员应当理解:其依然可以对前述各实施例所记载的技术方案进行修改,或者对其中部分或者全部技术特征进行等同替换;而这些修改或者替换,并不使相应技术方案的本质脱离本发明各实施例技术方案的范围,其均应涵盖在本发明的权利要求和说明书的范围当中。
Claims (9)
- 一种基于心电心音的儿童心血管疾病诊断方法,其特征在于,包括以下步骤:S1:获取待诊断患儿的心电信号和心音信号;S2:对所述心电信号和心音信号进行处理,得到多维心音特征;S3:将所述多维心音特征输入诊断模型,得到诊断结果,其中,所述诊断结果为心血管疾病类型。
- 根据权利要求1所述的一种基于心电心音的儿童心血管疾病诊断方法,其特征在于,S2中,对心电信号和心音信号进行处理,得到心电心音特征的方法包括:S21:对心电信号和心音信号进行预处理;S22:根据预处理后的心电信号和心音信号,获取初始心音特征;S23:通过时域、频域、能量域分析方法对所述初始心音特征进行分析,得到多维心音特征,所述多维心音特征包括时域心音特征、频域心音特征和能量域心音特征。
- 根据权利要求2所述的一种基于心电心音的儿童心血管疾病诊断方法,其特征在于,S21中,对心电信号和心音信号进行预处理的方法包括:S211:对心电信号和心音信号进行截取;S212:对采样后的心电信号和心音信号进行滤波降噪处理;S213:对滤波降噪后的心电信号和心音信号进行归一化处理。
- 根据权利要求2所述的一种基于心电心音的儿童心血管疾病诊断方法, 其特征在于,S22中,根据预处理后的心电信号和心音信号,获取初始心音特征的方法包括:S221:识别预处理后的心电信号的QRS波;S222:根据QRS波的Q波,从预处理后的心音信号中识别出第一心音的持续和幅值信息,根据所述QRS波的T波,从预处理后的心音信号中识别出第二心音的持续时间和幅值信息;S222:根据第一心音和第二心音的持续时间和幅值信息,获取收缩期的持续时间和幅值信息、舒张期的持续时间和幅值信息;S223:将所述第一心音、第二心音、收缩期和舒张期的持续时间和幅值信息作为初始心音特征。
- 根据权利要求2所述的一种基于心电心音的儿童心血管疾病诊断方法,其特征在于,S23中,通过时域分析方法对所述初始心音特征进行分析,得到时域心音特征的方法包括:分别计算第一心音、收缩期、第二心音和舒张期两两之间的幅值比;分别计算相邻第一心音、收缩期、第二心音和舒张期之间的时间差;计算EMAT%;计算各个幅值比、各个时间差、EMAT%在整个心动周期的均值,得到11个时域心音特征。
- 根据权利要求4所述的一种基于心电心音的儿童心血管疾病诊断方法,其特征在于,S23中,通过频域分析方法对所述初始心音特征进行分析,得到时域心音特征的方法包括:对所述初始心音特征进行傅里叶变换,得到对应的频谱序列,其中,所述频谱序列包括第一心音频谱序列、第二心音频谱序列、收缩期频谱序列和舒张期频谱序列;分别提取第一心音频谱序列、第二心音频谱序列、收缩期频谱序列和舒张期频谱序列中频率为多个指定频率的数值,求取各个数值在所有心动周期的均值,并形成新的频谱序列;得到4个频域心音特征。
- 根据权利要求6所述的一种基于心电心音的儿童心血管疾病诊断方法,其特征在于,所述第一心音频谱序列、第二心音频谱序列、收缩期、舒张期对应的多个指定频率为60HZ、70HZ、80HZ、90HZ、100HZ、110HZ、120HZ、130HZ、80HZ、150HZ。
- 根据权利要求4所述的一种基于心电心音的儿童心血管疾病诊断方法,其特征在于,S23中,通过频域分析方法对所述初始心音特征进行分析,得到时域心音特征的方法包括:分别计算第一心音、收缩期、第二心音和舒张期两两之间的能量比,并求各个能量比在所有心动周期的均值,得到6个能量域心音特征。
- 根据权利要求1所述的一种基于心电心音的儿童心血管疾病诊断方法,其特征在于,S3中,所述诊断模型包括输入层、隐藏层、输出层和softmax函数,其中,所述输入层由21个神经元组成,所述21个神经元对应多维心音特征;所述隐藏层分别有7和8个神经元;所述输出层由个8神经元组成,对应心血管疾病的类型;最后使用一个softmax函数,用于解决多分类问题而创建。
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
PCT/CN2022/103931 WO2024007152A1 (zh) | 2022-07-05 | 2022-07-05 | 一种基于心电心音的儿童心血管疾病诊断方法 |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
PCT/CN2022/103931 WO2024007152A1 (zh) | 2022-07-05 | 2022-07-05 | 一种基于心电心音的儿童心血管疾病诊断方法 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2024007152A1 true WO2024007152A1 (zh) | 2024-01-11 |
Family
ID=89454721
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/CN2022/103931 WO2024007152A1 (zh) | 2022-07-05 | 2022-07-05 | 一种基于心电心音的儿童心血管疾病诊断方法 |
Country Status (1)
Country | Link |
---|---|
WO (1) | WO2024007152A1 (zh) |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20050222515A1 (en) * | 2004-02-23 | 2005-10-06 | Biosignetics Corporation | Cardiovascular sound signature: method, process and format |
US20090192401A1 (en) * | 2008-01-25 | 2009-07-30 | Sourabh Ravindran | Method and system for heart sound identification |
US20130226019A1 (en) * | 2010-08-25 | 2013-08-29 | Diacoustic Medical Devices (Pty) Ltd | System and method for classifying a heart sound |
US20140276132A1 (en) * | 2013-03-15 | 2014-09-18 | Andreas J. Schriefl | Automated Diagnosis-Assisting Medical Devices Utilizing Rate/Frequency Estimation |
CN111329508A (zh) * | 2020-03-02 | 2020-06-26 | 浙江大学 | 用于先心病筛查的心脏杂音智能分析方法 |
CN112971839A (zh) * | 2021-02-05 | 2021-06-18 | 云南大学 | 一种基于前馈卷积神经网络的心音分类方法 |
CN115040135A (zh) * | 2022-06-09 | 2022-09-13 | 张福伟 | 一种基于心电心音的儿童心血管疾病诊断方法 |
-
2022
- 2022-07-05 WO PCT/CN2022/103931 patent/WO2024007152A1/zh unknown
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20050222515A1 (en) * | 2004-02-23 | 2005-10-06 | Biosignetics Corporation | Cardiovascular sound signature: method, process and format |
US20090192401A1 (en) * | 2008-01-25 | 2009-07-30 | Sourabh Ravindran | Method and system for heart sound identification |
US20130226019A1 (en) * | 2010-08-25 | 2013-08-29 | Diacoustic Medical Devices (Pty) Ltd | System and method for classifying a heart sound |
US20140276132A1 (en) * | 2013-03-15 | 2014-09-18 | Andreas J. Schriefl | Automated Diagnosis-Assisting Medical Devices Utilizing Rate/Frequency Estimation |
CN111329508A (zh) * | 2020-03-02 | 2020-06-26 | 浙江大学 | 用于先心病筛查的心脏杂音智能分析方法 |
CN112971839A (zh) * | 2021-02-05 | 2021-06-18 | 云南大学 | 一种基于前馈卷积神经网络的心音分类方法 |
CN115040135A (zh) * | 2022-06-09 | 2022-09-13 | 张福伟 | 一种基于心电心音的儿童心血管疾病诊断方法 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Leung et al. | Classification of heart sounds using time-frequency method and artificial neural networks | |
Syed et al. | A framework for the analysis of acoustical cardiac signals | |
Myint et al. | An electronic stethoscope with diagnosis capability | |
US20040260188A1 (en) | Automated auscultation system | |
Choi et al. | Selection of wavelet packet measures for insufficiency murmur identification | |
JPH021217A (ja) | 心疾患検出装置 | |
Javed et al. | A signal processing module for the analysis of heart sounds and heart murmurs | |
CN112971839B (zh) | 一种基于前馈卷积神经网络的心音分类方法 | |
Sedighian et al. | Pediatric heart sound segmentation using Hidden Markov Model | |
Sinha et al. | Backpropagation artificial neural network classifier to detect changes in heart sound due to mitral valve regurgitation | |
Wang et al. | Temporal-framing adaptive network for heart sound segmentation without prior knowledge of state duration | |
Banerjee et al. | Segmentation and detection of first and second heart sounds (Si and S 2) using variational mode decomposition | |
CN114469124A (zh) | 一种运动过程中异常心电信号的识别方法 | |
Song et al. | Heart sounds monitor and analysis in noisy environments | |
Akbari et al. | Digital Subtraction Phonocardiography (DSP) applied to the detection and characterization of heart murmurs | |
Deperlioglu | Segmentation of heart sounds by re-sampled signal energy method | |
CN111329508A (zh) | 用于先心病筛查的心脏杂音智能分析方法 | |
CN115040135A (zh) | 一种基于心电心音的儿童心血管疾病诊断方法 | |
Nizam et al. | Hilbert-envelope features for cardiac disease classification from noisy phonocardiograms | |
WO2024007152A1 (zh) | 一种基于心电心音的儿童心血管疾病诊断方法 | |
Botha et al. | Autonomous auscultation of the human heart employing a precordial electro-phonocardiogram and ensemble empirical mode decomposition | |
CN111528900A (zh) | 基于巴特沃斯滤波器与香农熵法的心音分段方法和装置 | |
Golpaygani et al. | Detection and identification of S1 and S2 heart sounds using wavelet decomposition method | |
Behbahani | A hybrid algorithm for heart sounds segmentation based on phonocardiogram | |
Ari et al. | On a robust algorithm for heart sound segmentation |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 22949740 Country of ref document: EP Kind code of ref document: A1 |