技术领域technical field
本发明涉及冠心病筛查设备技术领域，特别涉及一种冠心病筛查装置、筛 查系统以及信号特征提取方法。The present invention relates to the technical field of coronary heart disease screening equipment, in particular to a coronary heart disease screening device, a screening system and a signal feature extraction method.
背景技术Background technique
冠状动脉疾病(Coronary Artery Disease，缩写为CAD，简称冠心病)是全 球致死率最高的疾病之一。冠心病的基本病理过程是冠脉血管壁的增厚、变硬 以及斑块的沉积(包括脂肪、胆固醇、纤维等)，即冠状动脉粥样硬化。斑块的 沉积会使血管壁变窄，从而限制冠脉将富氧的血液输送至心肌。心肌供血不足 可能会引发胸痛、心肌梗死、心力衰竭或者心率不齐。相当一部分病人在猝死 前并没有任何明显的症状，因此对于冠心病的早期筛查尤为重要。Coronary artery disease (Coronary Artery Disease, abbreviated as CAD, referred to as coronary heart disease) is one of the diseases with the highest mortality rate in the world. The basic pathological process of coronary heart disease is the thickening and hardening of the coronary vessel wall and the deposition of plaque (including fat, cholesterol, fiber, etc.), that is, coronary atherosclerosis. Plaque deposits narrow the walls of blood vessels, limiting the ability of the coronary arteries to deliver oxygen-rich blood to the heart muscle. Insufficient blood flow to the heart muscle may cause chest pain, myocardial infarction, heart failure, or irregular heartbeat. A considerable number of patients do not have any obvious symptoms before sudden death, so early screening for coronary heart disease is particularly important.
目前用于冠心病的检测的方法有许多，其中冠状动脉造影是公认的“金标 准”。但由于冠状动脉造影需要用导管将造影剂送入冠脉，对医疗机构的设备和 医疗工作者的技术有很高的要求，而且冠状动脉造影价格昂贵，对病患带来创 伤，因此很难将该技术普及。因此，开发一种非侵入式、操作简便且价格低廉 的冠心病筛查方法是很有必要的。There are many methods currently used for the detection of coronary heart disease, among which coronary angiography is the recognized "gold standard". However, because coronary angiography needs to use a catheter to send a contrast agent into the coronary artery, it has high requirements for the equipment of medical institutions and the skills of medical workers, and coronary angiography is expensive and traumatic to patients, so it is difficult popularize the technology. Therefore, it is necessary to develop a non-invasive, easy-to-operate and low-cost screening method for coronary heart disease.
专利文献CN103841882A，公开了“一种用于检测冠状动脉疾病的系统， 其包括检测装置，所述检测装置包括：检测装置本体，至少一个连接至所述检 测装置本体的传感器，存储器，控制器，其配置为指令所述至少一个传感器进 行数据采样并且将采样数据存储在所述存储器中，和接触部，其配置为接触患 者；以及识别元件，所述识别元件包括至少一个与所述患者身上的数据获取位 置相对应的识别区域，所述至少一个识别区域配置为与所述接触位置连接，其 中所述系统配置为根据在所述与所述至少一个识别区域相关联的所述数据获取 位置由所述至少一个传感器采样的数据判定是否存在冠状动脉疾病”，该技术方 案仅仅考虑了心音检测在非侵入式冠心病筛查领域的应用，由于心音检测固有 的缺陷，筛查准确率不高，使用效果并不理想。Patent document CN103841882A discloses "a system for detecting coronary artery disease, which includes a detection device, and the detection device includes: a detection device body, at least one sensor connected to the detection device body, a memory, a controller, It is configured to instruct the at least one sensor to perform data sampling and store the sampled data in the memory, and a contact portion configured to contact a patient; and an identification element comprising at least one an identification area corresponding to a data acquisition location, the at least one identification area is configured to be connected to the contact location, wherein the system is configured to The data sampled by the at least one sensor determines whether there is coronary artery disease". This technical solution only considers the application of heart sound detection in the field of non-invasive coronary heart disease screening. Due to the inherent defects of heart sound detection, the screening accuracy is not high. The effect is not ideal.
然而，在现有的身体健康普查中，除了心音，脉搏波和心电信号也是可能 涉及的检测项目。其中，However, in the existing physical health survey, in addition to heart sounds, pulse waves and ECG signals are also possible detection items involved. in,
脉搏波是心脏的搏动沿着动脉血管和血流传播形成的，脉搏波传导速度是 指脉搏波在单位时间内沿动脉壁传导的距离，主要和传输介质的物理与几何性 质有关。当动脉血管的弹性降低(硬度增加)时，脉搏波的传导速度增加。动 脉血管的主要功能是输送血液和缓冲血液，其中弹性血管在缓冲血液和抑制压 力振荡方面起主导作用，并将搏动的血流转换为稳定的血流输送至全身，防止 心脏收缩产生的机械应力对动脉血管壁和微血管造成损伤。研究表明，测量脉 搏波传导速度可以评估动脉硬化程度，并间接反映冠脉的血流供应情况。此外， 粥样硬化往往并非局部病变，患有主动脉粥样硬化的患者往往其冠脉也可能存 在粥样硬化病变。因此，脉搏波传导速度与冠状动脉狭窄之间存在相关性，其 可以作为筛查冠心病的一个指标。The pulse wave is formed by the pulse of the heart propagating along the arterial vessels and blood flow. The pulse wave propagation velocity refers to the distance that the pulse wave travels along the arterial wall per unit time, which is mainly related to the physical and geometric properties of the transmission medium. When the elasticity of the arterial vessel decreases (hardness increases), the conduction velocity of the pulse wave increases. The main function of arterial blood vessels is to transport blood and buffer blood, in which elastic blood vessels play a leading role in buffering blood and suppressing pressure oscillations, transforming the pulsating blood flow into a stable blood flow and delivering it to the whole body, preventing the mechanical stress generated by heart contraction Damage to arterial walls and capillaries. Studies have shown that measuring pulse wave velocity can evaluate the degree of arterial stiffness and indirectly reflect the blood supply of coronary arteries. In addition, atherosclerosis is often not a local lesion, and patients with aortic atherosclerosis often have atherosclerotic lesions in their coronary arteries. Therefore, there is a correlation between pulse wave velocity and coronary artery stenosis, which can be used as an index for screening coronary heart disease.
心电信号反映的是心脏在每个心动周期中，由起搏点、心房、心室相继兴 奋，并伴随有心肌细胞膜电位的变化。临床上，心电波形的变化可以反映心脏 的功能异常。例如心肌梗塞病人的心电波形有较长的QT间期和异常的高幅值T 波，而心内膜心肌缺血和大面积缺血病人的ST段的幅值则分别低于和高于正 常人。QT间期是心电图波段中，心室除极到完全复极的时间。T波在心电图波 段中表示心室复极化。ST段在心电图波段中表示心室除极完成。ECG signals reflect that the heart is excited successively by the pacemaker, atrium, and ventricle in each cardiac cycle, accompanied by changes in the membrane potential of myocardial cells. Clinically, changes in the ECG waveform can reflect abnormal heart function. For example, the ECG waveforms of patients with myocardial infarction have longer QT intervals and abnormally high-amplitude T waves, while the amplitudes of ST segments in patients with endomyocardial ischemia and massive ischemia are lower and higher, respectively. normal people. The QT interval is the time between ventricular depolarization and complete repolarization in the ECG wave band. The T wave represents ventricular repolarization in the ECG spectrum. The ST segment represents the completion of ventricular depolarization on the ECG wave.
由于心电波形中夹杂有许多噪声，并且异常波形的出现没有规律，因此仅 凭肉眼很难对冠心病做出诊断。目前已经有研究将机器学习用于ECG信号的分 析。此外，心电信号还可以用于校正脉搏波的波形。Because there is a lot of noise in the ECG waveform, and the appearance of abnormal waveforms is irregular, it is difficult to diagnose coronary heart disease with the naked eye. There have been studies using machine learning for the analysis of ECG signals. In addition, the ECG signal can also be used to correct the waveform of the pulse wave.
本文中的关于心音信号的分割参考了文献Segmentation of heart soundrecordings by a duration-dependent hidden Markov model.PhysiologicalMeasurement,2010.31(4):p.513-529.The segmentation of heart sound signals in this article refers to the literature Segmentation of heart sound recordings by a duration-dependent hidden Markov model.PhysiologicalMeasurement,2010.31(4):p.513-529.
本文中关于心电信号的处参考了以下文献：The following literature is referred to in this article about the ECG signal:
熊敏and刘雄飞,基于多孔算法的心电图QRS波检测.计算机仿真,2011. 28(12):p.244-248.Xiong Min and Liu Xiongfei, Electrocardiogram QRS Wave Detection Based on Porous Algorithm. Computer Simulation, 2011. 28(12): p.244-248.
Illanes-Manriquez,A.and Q.Zhang.An algorithm for robust detection ofQRS onset and offset in ECG signals.in Computers in Cardiology.2013.Illanes-Manriquez, A. and Q. Zhang. An algorithm for robust detection of QRS onset and offset in ECG signals. in Computers in Cardiology. 2013.
芦继来and胡广书,基于小波变换的运动心电ST段检测方法.北京生物医学 工程,2005.24(5):p.329-332.Lu Jilai and Hu Guangshu, ST segment detection method of exercise ECG based on wavelet transform. Beijing Biomedical Engineering, 2005.24(5): p.329-332.
发明内容Contents of the invention
本发明的实施例公开了一种冠心病筛查装置、筛查系统以及信号特征提取 方法，目的在于解决现有的冠心病筛查系统准确率偏低的问题。The embodiment of the present invention discloses a coronary heart disease screening device, a screening system and a signal feature extraction method, aiming to solve the problem of low accuracy of the existing coronary heart disease screening system.
一个实施例中，一种冠心病筛查装置，所述筛查装置包括：拾音器，用于 获取心音信号。脉搏波传感器，用于获取脉搏波信号。心电传感器，用于获取 心电信号。微处理器，连接拾音器、脉搏波传感器和心电传感器的输出端，获 取心音信号、脉搏波信号和心电信号，经过分析判断，给出冠心病的筛查结果。 筛查诊断装置还包括与微处理器连接的存储器，该存储器用于存储程序指令、 用于诊断的原始数据和/或筛查诊断结果。In one embodiment, a screening device for coronary heart disease, the screening device includes: a sound pickup for acquiring heart sound signals. The pulse wave sensor is used to obtain the pulse wave signal. The ECG sensor is used to obtain the ECG signal. The microprocessor is connected to the output terminals of the pickup, the pulse wave sensor and the ECG sensor to obtain the heart sound signal, the pulse wave signal and the ECG signal, and after analyzing and judging, it gives the screening result of coronary heart disease. The screening and diagnosing device also includes a memory connected to the microprocessor for storing program instructions, raw data for diagnosis and/or screening and diagnosing results.
另一实施例中，一种冠心病筛查系统，包括：拾音器，用于获取心音信号； 脉搏波传感器，用于获取脉搏波信号；心电传感器，用于获取心电信号；存储 器以及处理器。处理器执行操作包括：获取拾音器的心音信号、脉搏波传感器 的脉搏波信号和心电传感器的心电信号。将从心音、脉搏波和心电数据中获取 的特征与用户病史资料和用户基本生理参数进行结合，获得特征向量。将所述 特征向量输入诊断模型，获得筛查结果。诊断模型基于径向基函数(RBF)神 经网络。将所述的筛查结果以及对该筛查诊断结果的验证结果存入数据库，并 且作为训练样本，采用最近邻聚类学习算法，构建径向基函数(RBF)神经网 络。In another embodiment, a coronary heart disease screening system includes: a sound pick-up for obtaining heart sound signals; a pulse wave sensor for obtaining pulse wave signals; an electrocardiographic sensor for obtaining electrocardiographic signals; a memory and a processor . The operations performed by the processor include: acquiring the heart sound signal of the pickup, the pulse wave signal of the pulse wave sensor and the ECG signal of the ECG sensor. Combine the features obtained from heart sound, pulse wave and ECG data with the user's medical history data and the user's basic physiological parameters to obtain the feature vector. The feature vector is input into the diagnosis model to obtain the screening result. The diagnostic model is based on a radial basis function (RBF) neural network. The screening results and the verification results of the screening diagnosis results are stored in the database, and as training samples, the nearest neighbor clustering learning algorithm is used to construct a radial basis function (RBF) neural network.
又一个实施例中，一种信号特征提取方法，所述信号包括获取的心电信号、 心音信号和脉搏波信号，对信号特征的提取方法包括以下步骤：In yet another embodiment, a method for extracting signal features, the signal includes acquired ECG signals, heart sound signals and pulse wave signals, and the method for extracting signal features includes the following steps:
对于心电信号，确定其ST段电平和QRS波群宽度；For the ECG signal, determine its ST segment level and QRS complex width;
对于心音信号，确定其频谱质心、频谱滚降点及通过小波变换得到的5个 频率区间的舒张期能量与整个心动周期能量的比值；For the heart sound signal, determine its spectral centroid, spectral roll-off point and the diastolic energy of 5 frequency intervals obtained by wavelet transform and the ratio of the whole cardiac cycle energy;
对于脉搏波信号，确定踝肱脉搏波传导速度(baPWV)和踝肱指数(ABI)。For the pulse wave signal, the ankle-brachial pulse wave velocity (baPWV) and the ankle-brachial index (ABI) were determined.
本发明的有益效果包括：The beneficial effects of the present invention include:
1.通过对病患的心音、脉搏波和心电信号三者的结合采集和分析，提高了 冠心病筛查的准确性。1. Through the combined acquisition and analysis of the patient's heart sound, pulse wave and ECG signal, the accuracy of coronary heart disease screening is improved.
2.通过进一步综合考虑病患的生理指标，提高冠心病筛查的准确度。2. By further comprehensively considering the physiological indicators of patients, the accuracy of coronary heart disease screening can be improved.
3.通过建立病患的冠心病筛查模型，利用海量的训练样本，提高了冠心病 筛查的智能化程度。3. By establishing a coronary heart disease screening model for patients and using a large number of training samples, the intelligence of coronary heart disease screening is improved.
4.通过引入径向基函数(RBF)神经网络，采用深度学习技术，不断优化 冠心病的筛查模型的准确度，达到了接近冠脉造影的判断水平。4. By introducing radial basis function (RBF) neural network and adopting deep learning technology, the accuracy of the screening model for coronary heart disease is continuously optimized, reaching a judgment level close to that of coronary angiography.
附图说明Description of drawings
通过参考附图阅读下文的详细描述，本发明示例性实施方式的上述以及其他目的、特征和优点将变得易于理解。在附图中，以示例性而非限制性的方式示出 了本发明的若干实施方式，其中：The above and other objects, features and advantages of exemplary embodiments of the present invention will become readily understood by reading the following detailed description with reference to the accompanying drawings. In the drawings, several embodiments of the invention are shown by way of illustration and not limitation, in which:
图1本发明实施例中冠脉疾病筛查的基本流程图。Fig. 1 is a basic flowchart of coronary artery disease screening in the embodiment of the present invention.
图2本发明实施例中冠心病筛查装置的基本结构图。Fig. 2 is a basic structural diagram of the coronary heart disease screening device in the embodiment of the present invention.
图3本发明实施例中RBF神经网络结构示意图。Fig. 3 is a schematic diagram of the structure of the RBF neural network in the embodiment of the present invention.
具体实施方式Detailed ways
根据一个或多个实施例，如图2所示，一种冠心病筛查装置，该装置包括： 用于获取心音信号的拾音器，用于获取脉搏波信号的脉搏波传感器，和用于获 取心电信号的心电传感器。还包括微处理器，连接拾音器、脉搏波传感器和心 电传感器的输出端，获取心音信号、脉搏波信号和心电信号，经过信号处理与 机器学习，给出冠心病的筛查结果。According to one or more embodiments, as shown in Figure 2, a coronary heart disease screening device, the device includes: a pickup for acquiring heart sound signals, a pulse wave sensor for acquiring pulse wave signals, and a pulse wave sensor for acquiring heart sound signals. ECG sensor for electrical signal. It also includes a microprocessor, connected to the output terminals of the pickup, pulse wave sensor and ECG sensor, to obtain heart sound signals, pulse wave signals and ECG signals, and after signal processing and machine learning, the screening results for coronary heart disease are given.
根据一个或多个实施例，拾音器、脉搏波传感器和心电传感器的输出端分别 经由滤波器和放大器的处理接入所述微处理器的输入端。筛查装置还包括与微 处理器连接的存储器，该存储器用于存储程序指令、用于诊断的原始数据和/ 或筛查结果。筛查装置还包括显示器，该显示器用于显示拾音器、脉搏波传感 器和心电传感器的实时信号，以及冠心病筛查结果。根据一个或多个实施例， 存储器包括计算机可读记录/存储介质，如随机存取存储器(RAM)、只读存储 器(ROM)、闪存存储器、光盘、磁盘、固态盘等等。According to one or more embodiments, the output terminals of the pickup, the pulse wave sensor and the electrocardiogram sensor are respectively connected to the input terminal of the microprocessor through the processing of the filter and the amplifier. The screening device also includes a memory coupled to the microprocessor for storing program instructions, raw data for diagnosis and/or screening results. The screening device also includes a display, which is used to display the real-time signals of the pickup, the pulse wave sensor and the electrocardiogram sensor, as well as the screening result of coronary heart disease. According to one or more embodiments, the memory includes computer readable recording/storage media such as random access memory (RAM), read only memory (ROM), flash memory, optical disks, magnetic disks, solid state disks, and the like.
根据一个或多个实施例，所述筛查装置还包括无线通讯模块，该模块用于 通信连接远程服务器和/或PC机，用于将诊断数据或筛查结果发送至远程服务 器，或者同时接收PC机给所述筛查装置的控制指令。According to one or more embodiments, the screening device further includes a wireless communication module, which is used for communicating with a remote server and/or a PC, for sending diagnostic data or screening results to a remote server, or simultaneously receiving The PC gives control instructions to the screening device.
根据一个或多个实施例，拾音器、脉搏波传感器和心电传感器的输出端分 别经由滤波器、放大器和模数转换器的处理接入所述微处理器的输入端。According to one or more embodiments, the output terminals of the pickup, the pulse wave sensor and the electrocardiogram sensor are respectively connected to the input terminal of the microprocessor through the processing of the filter, the amplifier and the analog-to-digital converter.
根据一个或多个实施例，如图1所示，微处理器被配置为执行存储在存储 器中的指令或执行与微处理器连接的存储器中的指令，微处理器执行的操作包 括：According to one or more embodiments, as shown in Figure 1, the microprocessor is configured to execute instructions stored in the memory or to execute instructions in the memory connected to the microprocessor, and the operations performed by the microprocessor include:
对从拾音器、脉搏波传感器和心电传感器获取的信号进行数据分析和处理；Perform data analysis and processing on the signals obtained from the pickup, pulse wave sensor and ECG sensor;
将心音、脉搏波和心电数据结合用户病史资料和用户基本生理参数进行特征 提取，获得特征向量；Combine the heart sound, pulse wave and ECG data with the user's medical history data and the user's basic physiological parameters for feature extraction to obtain the feature vector;
将所述特征向量输入筛查模型，获得筛查结果。Input the feature vector into the screening model to obtain the screening result.
其中，所述的对从拾音器、脉搏波传感器和心电传感器获取的信号进行数 据分析和处理包括：Wherein, the described data analysis and processing of the signals obtained from the pickup, the pulse wave sensor and the electrocardiogram sensor include:
对获取的心音信号进行分割处理、频谱分析和小波变换；Carry out segmentation processing, spectrum analysis and wavelet transform on the obtained heart sound signal;
计算踝肱脉搏波速度(baPWV)和踝肱指数(ABI)；Calculation of ankle-brachial pulse wave velocity (baPWV) and ankle-brachial index (ABI);
计算心电信号的ST段电平和QRS波群宽度。Calculate ST segment level and QRS wave group width of ECG signal.
将所述的对该筛查结果的验证结果上传数据库，并加入训练样本集，反复 训练神经网络，进一步提高模型对冠心病还筛查的准确度。The verification results of the screening results are uploaded to the database, and added to the training sample set, and the neural network is repeatedly trained to further improve the accuracy of the model for coronary heart disease screening.
根据一个或多个实施例，所述筛查模型的建立方法包括：According to one or more embodiments, the method for establishing the screening model includes:
对样本特征的确定，具体有：The determination of sample characteristics includes:
对于心电信号，需要确定ST段电平和QRS波群宽度，For the ECG signal, it is necessary to determine the ST segment level and the QRS complex width,
对于心音信号，需要确定其频谱质心、频谱滚降点及通过小波变换得到 的5个频率区间的舒张期功率与整个心动周期能量的比值。For the heart sound signal, it is necessary to determine its spectral centroid, spectral roll-off point and the ratio of the diastolic power of the five frequency intervals obtained by wavelet transform to the energy of the entire cardiac cycle.
对于脉搏波信号，需要确定踝肱脉搏波传导速度(baPWV)和踝肱指 数(ABI)，For the pulse wave signal, the ankle-brachial pulse wave velocity (baPWV) and the ankle-brachial index (ABI) need to be determined,
基本生理参数包括与血液检测指标有关的空腹血糖、甘油三酯、总胆固 醇、高密度脂蛋白胆固醇、低密度脂蛋白胆固醇、脂蛋白α，糖化血红蛋 白、以及用户的年龄、性别和/或BMI指数中的一种或任意组合。Basic physiological parameters include fasting blood glucose, triglycerides, total cholesterol, high-density lipoprotein cholesterol, low-density lipoprotein cholesterol, lipoprotein alpha, glycated hemoglobin, and user's age, gender and/or BMI index related to blood test indicators one or any combination of them.
病史类型参数包括：冠心病史、高血压史、糖尿病史、吸烟史、饮酒 史、高血脂症史、PCI手术史、心脏旁路移植手术史中的一种或者任意组合。Medical history type parameters include: history of coronary heart disease, history of hypertension, history of diabetes, history of smoking, history of drinking, history of hyperlipidemia, history of PCI surgery, history of heart bypass transplantation, or any combination thereof.
挑选一批具有代表性的数据(包括病情各异的冠心病患者和正常人)组成 训练样本，由经验丰富的医生根据训练样本的冠脉造影图像对每个训练样本进 行SYNTAX评分，该评分用于表征样本的冠状动脉的狭窄程度使用训练样本，A batch of representative data (including patients with coronary heart disease and normal people with different conditions) is selected to form training samples, and experienced doctors perform SYNTAX scores on each training sample according to the coronary angiography images of the training samples. The score is used Using training samples to characterize the degree of stenosis of the coronary arteries of the samples,
采用最近邻聚类学习算法，构建径向基函数(RBF)神经网络。The radial basis function (RBF) neural network is constructed by using the nearest neighbor clustering learning algorithm.
以下，对于该实施例中利用径向基函数神经网络筛查冠心病的方法具体描 述。Below, the method for utilizing radial basis function neural network screening for coronary heart disease in this embodiment is described in detail.
1、特征向量的筛选包括1. The screening of feature vectors includes
1.1心电信号的特征提取1.1 Feature extraction of ECG signal
1.1.1 QRS波群宽度检测1.1.1 QRS complex width detection
选取三次B样条小波作为基小波，运用其高阶导数对干扰信号进行平滑， 利用其特有的高阶平滑特性最大限度去除低频和高频干扰。The cubic B-spline wavelet is selected as the base wavelet, and its high-order derivative is used to smooth the interference signal, and its unique high-order smoothing characteristics are used to remove low-frequency and high-frequency interference to the greatest extent.
对信号进行三次B样条小波分解，共分为五个尺度，并在第三个尺度找到 所有的模极大值点，指定阈值区间，保留在阈值区间内的模极大值点。在模极 大值点中找出模极大值对，提取模极大值对的零点交叉点，并进行时移修正， 得到R波峰值点tp。Carry out cubic B-spline wavelet decomposition on the signal, divide it into five scales, and find all the modulus maximum points in the third scale, specify the threshold interval, and keep the modulus maximum points within the threshold interval. Find the modulus maximum pair from the modulus maximum point, extract the zero crossing point of the modulus maximum pair, and perform time shift correction to obtain the R wave peak point tp .
从R波峰值点tp向前300ms，向后200ms分别划定两个区间，在这两个区 间中分别检测QRS波群的起点和终点。下面以QRS波群终点的检测为例，QRS 波群起点的检测采用相同的步骤。From the R wave peak pointtp , 300 ms forward and 200 ms backward are respectively delineated into two intervals, and the starting point and end point of the QRS complex are detected in these two intervals. Taking the detection of the end point of the QRS complex as an example below, the detection of the starting point of the QRS complex adopts the same steps.
①对信号进行希尔伯特变换得到QRS波群的包络，用env(t)表示；① Perform Hilbert transform on the signal to obtain the envelope of the QRS complex, represented by env(t);
②用一个滑动的矩形窗R在区间内移动，计算该窗与QRS包络相交的区间 内的面积A(t)② Use a sliding rectangular window R to move in the interval, and calculate the area A(t) in the interval where the window intersects with the QRS envelope
其中W为窗口的宽度where W is the width of the window
③设定窗口的初始宽度为W0，即实际情况下QRS波群可能的最大宽度。移 动该窗口并计算A(t)，当A(t)最大时的窗口最右端的位置记为s2；③ The initial width of the window is set to W0 , which is the maximum possible width of the QRS complex in actual conditions. Move the window and calculate A(t), when A(t) is the largest, the rightmost position of the window is recorded as s2 ;
④设定窗口的实际宽度W＝s2-tp。再次移动该窗口并计算A(t)，记下当 A(t)值最大时的窗口最右端的位置，该点即为QRS波群的终点。④ Set the actual width of the window W=s2 -tp . Move the window again and calculate A(t), and record the position of the rightmost end of the window when the value of A(t) is maximum, which is the end point of the QRS complex.
将QRS波群的终点和起点的时间坐标相减，即可得到QRS波群的宽度， 同一病人多个QRS波群宽度取平均值(单位：ms)。The width of the QRS complex can be obtained by subtracting the time coordinates of the end point and the starting point of the QRS complex, and the average value (unit: ms) of multiple QRS complex widths of the same patient is taken.
1.1.2 ST段幅值检测1.1.2 ST segment amplitude detection
采用J+60/80方法确定ST段采样点位置。其中J点即为上文求得的QRS 波群的终点，在心率小于120bpm时，采用J+80，即将J点后80ms处作为ST 段的采样点；在心率大于120bpm时，采用J+60，即将J点后60ms处作为ST 段的采样点。为了提高可靠性，将采样点前后各8ms的数据的平均值作为ST 段的幅值，同一病人多个ST段幅值取平均值(单位：mV)。The J+60/80 method was used to determine the location of the ST segment sampling point. Among them, point J is the end point of the QRS complex obtained above. When the heart rate is less than 120bpm, use J+80, that is, 80ms after point J as the sampling point of ST segment; when the heart rate is greater than 120bpm, use J+60 , that is, 60ms after point J is taken as the sampling point of ST segment. In order to improve the reliability, the average value of the data of 8ms before and after the sampling point is taken as the amplitude of the ST segment, and the average value (unit: mV) of multiple ST segment amplitudes of the same patient is taken.
1.2心音信号特征提取，包括：1.2 Heart sound signal feature extraction, including:
1.2.1心音信号分割(基于持续时间依赖的隐马尔科夫模型)1.2.1 Heart sound signal segmentation (based on duration-dependent hidden Markov model)
首先从样本中取出一部分作为模型训练样本，并将这些样本手动分割为四 个区域：第一心音、静息收缩期、第二心音、静息舒张期，作为马尔科夫模型 的四个状态。对这些分割好的信号片段进行带通滤波(25Hz，1000Hz)。求取 第一心音和第二心音的包络。用包络的97百分位数对包络进行归一化以减少个 体之间的差异。First, take a part of the sample as a model training sample, and manually divide these samples into four regions: the first heart sound, resting systole, second heart sound, and resting diastole, as the four states of the Markov model . Perform bandpass filtering (25Hz, 1000Hz) on these segmented signal segments. Find the envelopes of the first heart sound and the second heart sound. Envelopes were normalized by their 97th percentile to reduce inter-individual variability.
在传统的隐马尔科夫模型的基础上为每一个状态补充一个状态持续时间的 概率分布函数Pj(d)即产生了状态持续时间依赖的隐马尔科夫模型。On the basis of the traditional hidden Markov model, a probability distribution function Pj (d) of the state duration is supplemented for each state, which produces a state duration-dependent hidden Markov model.
利用训练样本得到模型的参数——状态转移概率矩阵、初始状态概率向量 和四种状态关于持续时间的概率分布，然后用训练好的模型去处理未分割的心 音信号。Use the training samples to obtain the parameters of the model - the state transition probability matrix, the initial state probability vector and the probability distribution of the four states with respect to the duration, and then use the trained model to process the unsegmented heart sound signal.
保存使下式最大化的持续时间d和上一状态i,之后用回溯算法得到隐式 状态链，即每一个采样点从观测值分别属于哪个状态。Save the duration d and the previous state i that maximize the following formula, and then use the backtracking algorithm to obtain the implicit state chain, that is, which state each sampling point belongs to from the observed value.
上式中，δ(j)是指下一时刻从状态j转换到一个新的状态的概率，δt-d(i)是指 上一个状态i在时刻t-d结束的概率，aij是从状态i转变为状态j的概率，Pj(d)是 在状态中的关于状态j持续时间d的概率分布，是指从t-d到t这段 时间内状态j产生可观测序列的概率。In the above formula, δ(j) refers to the probability of transitioning from state j to a new state at the next moment, δtd (i) refers to the probability that the previous state i ends at time td, aij is the transition from state i is the probability of state j, Pj (d) is the probability distribution in the state with respect to state j duration d, It refers to the probability that state j produces an observable sequence during the period from td to t.
用上述的持续时间依赖的隐马尔科夫模型可以精确地将每个心音周期分为 第一心音、第二心音、静息收缩期和静息舒张期四段。Using the above-mentioned duration-dependent hidden Markov model, each heart sound cycle can be accurately divided into four segments: the first heart sound, the second heart sound, the resting systolic period and the resting diastolic period.
1.2.2频谱质心和频谱滚降点的计算1.2.2 Calculation of spectral centroid and spectral roll-off point
在求取频谱质心和频谱滚降点前需要对信号进行快速傅里叶变换。Before calculating the spectral centroid and spectral roll-off point, it is necessary to perform fast Fourier transform on the signal.
求取频谱质心公式如下：The formula for obtaining the spectral centroid is as follows:
其中，X(k)是对心音信号进行快速傅里叶变换后经求模运算得到的频谱序列， M是该序列的样本点数量，求取每个周期的频谱质心取平均值(单位：Hz)。Among them, X(k) is the frequency spectrum sequence obtained by the modulo operation after the fast Fourier transform of the heart sound signal, M is the number of sample points of the sequence, and the spectral centroid of each cycle is calculated to get the average value (unit: Hz ).
求取频谱滚降点公式如下：The formula for calculating the spectrum roll-off point is as follows:
fc即为频谱滚降点,求出同一病人多个周期的频谱滚降点取平均值，(单位： Hz)。fc is the spectrum roll-off point, and the average value of the spectrum roll-off points of multiple cycles of the same patient is calculated (unit: Hz).
1.2.3利用小波变换得到不同频率区间的舒张期能量和整周期能量的比值1.2.3 Using wavelet transform to obtain the ratio of diastolic energy and whole cycle energy in different frequency intervals
小波积分变换公式如下：The wavelet integral transformation formula is as follows:
其中，是基本小波，f(t)是信号。a是经变换后与频率有关的伸 缩因子，反比于频率；τ是信号变换后与时间相关的位移因子。小波变换 能够将不同频率成分的混合信号分解为不同频率块的信号，这对于分析本 身频率不高但包含有与疾病相关的高频成分的心音信号很有帮助。在这里 采用中心三次B样条小波，对信号(采样频率f＝4kHz)进行半周期离散 小波变换，当小波尺度为5时，心音信号频域被分为6个区域，分别是 1000-2000Hz，500-1000Hz，250-500Hz，125-250Hz，62.5-125Hz和残余信 号0-62.5Hz。计算前5个频率段的舒张期能量与完整周期能量的比值Q1-Q5， 同一病人多个周期的能量比值(Q1-Q5)分别取平均值，该值无量纲单位。in, is the basic wavelet, and f(t) is the signal. a is the expansion factor related to frequency after transformation, which is inversely proportional to frequency; τ is the displacement factor related to time after signal transformation. Wavelet transform can decompose mixed signals of different frequency components into signals of different frequency blocks, which is very helpful for analyzing heart sound signals that are not high in frequency but contain high frequency components related to diseases. Here, the central cubic B-spline wavelet is used to perform half-period discrete wavelet transform on the signal (sampling frequency f=4kHz). When the wavelet scale is 5, the frequency domain of the heart sound signal is divided into 6 regions, which are 1000-2000Hz respectively. 500-1000Hz, 250-500Hz, 125-250Hz, 62.5-125Hz and residual signal 0-62.5Hz. Calculate the ratio Q1-Q5 of the diastolic energy of the first 5 frequency segments to the energy of the complete cycle, and the energy ratios (Q1-Q5) of multiple cycles of the same patient are respectively averaged, and the value has no dimension unit.
信号能量：其中，N是信号中的样本总数，x[k]是离散 的信号序列。Signal energy: Among them, N is the total number of samples in the signal, and x[k] is the discrete signal sequence.
1.3脉搏波信号特征提取1.3 Pulse wave signal feature extraction
1.3.1踝肱脉搏波速度(baPWV)1.3.1 Ankle Brachial Pulse Wave Velocity (baPWV)
在肱动脉脉搏波波形和胫后动脉脉搏波波形中各选择一个特征点，利用相 交切线算法计算这两个点之间的时间差即为脉搏波的传导时间Δt。在体表测量 肱动脉处的测量点到胸骨上切迹间的距离记为Lb，在体表测量胫后动脉处的测 量点到胸骨上切迹之间的距离记为La。踝肱脉搏波传导速度求出左边的baPWV和右边的baPWV然后取平均值(单位：cm/s)。Select a characteristic point in the brachial artery pulse wave waveform and the posterior tibial artery pulse wave waveform, and calculate the time difference between these two points by using the intersection and tangent algorithm, which is the pulse wave transit time Δt . The distance between the measurement point at the brachial artery and the suprasternal notch was measured on the body surface as Lb , and the distance between the measurement point at the posterior tibial artery and the suprasternal notch was recorded as La . ankle brachial pulse wave velocity Find the baPWV on the left and the baPWV on the right and take the average value (unit: cm/s).
1.3.2踝肱指数(ABI)1.3.2 Ankle Brachial Index (ABI)
用示波压力传感器测得肱动脉和胫后动脉的收缩压，肱动脉收缩压取两侧 测量值的高值，并分别以左踝及右踝动脉的收缩压计算相应的左右两侧踝肱指 数，最后取左右两侧的ABI的低值作为该患者的ABI(无量纲单位)。The systolic pressure of the brachial artery and the posterior tibial artery was measured with an oscillometric pressure sensor. The systolic pressure of the brachial artery was the high value of the measured values on both sides, and the corresponding left and right ankle-brachial arteries were calculated based on the systolic pressure of the left and right ankle arteries. Index, and finally take the low value of the ABI on the left and right sides as the patient's ABI (dimensionless unit).
1.4用户的基本生理指标和病史1.4 Basic physiological indicators and medical history of users
1.4.1基本生理指标1.4.1 Basic physiological indicators
用户在医院进行以下血液指标的检测，包括：空腹血糖FPG(mmol/L)、甘 油三酯TG(mmol/L)、总胆固醇TC(mmol/L)、高密度脂蛋白胆固醇HDL-C(mmol/L)、低密度脂蛋白胆固醇LDL-C(mmol/L)、糖化血红蛋白HbA1c(％)和脂蛋白αLP(α)(g/L)；在测量踝肱脉搏波时可同时获取病人的血压指标，包 括：收缩压SBP(mmHg)、舒张压DBP(mmHg)和平均动脉压MAP(mmHg) (MAP＝(SBP+2*DBP)/3)；通过问卷调查或者体格检查得到用户的其他指标， 包括：年龄(岁)、BMI指数(kg/m2)和性别(男-1女-0)。The user tests the following blood indicators in the hospital, including: fasting blood glucose FPG(mmol/L) , triglyceride TG(mmol/L) , total cholesterol TC(mmol/L) , high-density lipoprotein cholesterol HDL-C(mmol /L) , low-density lipoprotein cholesterol LDL-C (mmol/L ), glycosylated hemoglobin HbA1c (%) and lipoprotein αLP (α) (g/L ); when measuring the ankle-brachial pulse wave, the patient's Blood pressure indicators, including: systolic blood pressure SBP (mmHg ), diastolic blood pressure DBP (mmHg ) and mean arterial pressure MAP (mmHg ) (MAP=(SBP+2*DBP)/3); other information obtained by the user through questionnaire survey or physical examination Indicators include: age (years), BMI index (kg/m2 ) and gender (male-1female-0).
1.4.2病史类型1.4.2 Types of medical history
通过问卷调查或者调取患者个人就诊记录得到病人的病史资料，包括：Obtain the patient's medical history information through questionnaire survey or access to the patient's personal medical records, including:
冠心病家族史(是-1，否-0) 糖尿病(是-1，否-0)Family history of coronary heart disease (yes-1, no-0) diabetes (yes-1, no-0)
吸烟(是-1，否-0) 饮酒(是-1，否-0)Smoking (Yes-1, No-0) Drinking (Yes-1, No-0)
高血脂(是-1，否-0) 经皮冠状动脉介入治疗史(是-1，否-0)Hyperlipidemia (Yes-1, No-0) History of percutaneous coronary intervention (Yes-1, No-0)
心脏旁路移植手术史(是-1，否-0)。History of cardiac bypass transplantation (yes-1, no-0).
2、训练样本的生成2. Generation of training samples
S Y N T A X积分是一种新的根据冠状动脉病变解剖特点进行危险分层的积 分系统,根据病变位置和严重程度、分叉、钙化等解剖特点定量评价冠状动脉病 变的复杂程度,以期指导治疗手段。SYNTAX积分将冠心病冠脉造影结果分为三 类。轻度狭窄(0-22分)、中度狭窄(23-32分)、重度狭窄(≥33分)。The SYNTAX score is a new scoring system for risk stratification based on the anatomical characteristics of coronary artery lesions. It quantitatively evaluates the complexity of coronary artery lesions based on anatomical characteristics such as lesion location and severity, bifurcations, and calcifications, in order to guide treatment methods. The SYNTAX score divides the coronary angiographic results of coronary heart disease into three categories. Mild stenosis (0-22 points), moderate stenosis (23-32 points), severe stenosis (≥33 points).
挑选一批具有代表性的数据(包括病情各异的冠心病患者和正常人)，组成 训练样本，由经验丰富的医生根据训练样本的冠脉造影图像对每个训练样本进行 SYNTAX评分作为样本的标签。Select a batch of representative data (including patients with coronary heart disease and normal people with different conditions) to form training samples, and experienced doctors perform SYNTAX scores on each training sample according to the coronary angiography images of the training samples as the sample Label.
3、筛查模型——径向基函数(RBF)神经网络3. Screening model - radial basis function (RBF) neural network
3.1RBF神经网络简介3.1 Introduction to RBF neural network
径向基函数(radial basis function，RBF)神经网络是一种前馈式神经网络， 具有最佳逼近和全局最优的性能，训练方法快速易行，且不存在局部最优的问 题，具有广泛的应用。The radial basis function (RBF) neural network is a feed-forward neural network, which has the best approximation and global optimal performance. The training method is fast and easy, and there is no local optimal problem. It has a wide range of Applications.
RBF网络通过径向基函数将数据非线性映射到一个高维的线性空间，然后 在高维空间用线性模型来做拟合或者回归。其网络结构如图3所示。该网络包 括三层，第一层为输入层，共N个节点(即特征)；第二层为隐藏层，共M个 节点，每个节点都是一个激活函数，用于将输入层的数据非线性映射到高维空 间；第三层为输出层，这里只输出一个值。在这里，RBF神经网络的输出便是 SYNTAX积分的预测值，根据网络输出评估病人的冠状动脉病变的严重程度。 径向基函数通常是高斯型的，如下所示：The RBF network maps the data nonlinearly to a high-dimensional linear space through the radial basis function, and then uses a linear model for fitting or regression in the high-dimensional space. Its network structure is shown in Figure 3. The network consists of three layers, the first layer is the input layer, a total of N nodes (ie features); the second layer is a hidden layer, a total of M nodes, each node is an activation function, used to convert the data of the input layer Non-linear mapping to high-dimensional space; the third layer is the output layer, where only one value is output. Here, the output of the RBF neural network is the predicted value of the SYNTAX score, and the severity of the patient's coronary artery lesion is evaluated according to the network output. Radial basis functions are usually Gaussian, as follows:
其中，Pj是隐藏层的第j个节点的输出，cj为第j个径向基函数的中心，x是 输入的特征向量、σj是隐藏层第j个节点的归一化参数。Among them, Pj is the output of the j-th node of the hidden layer, cj is the center of the j-th radial basis function, x is the input feature vector, and σj is the normalization parameter of the j-th node of the hidden layer.
3.2 RBF网络的构建——最近邻聚类学习算法3.2 Construction of RBF network - nearest neighbor clustering learning algorithm
按照RBF中心选取方式的不同，RBF神经网络的学习算法大致可以分为随 机选取中心、自组织选取中心、最近邻聚类法、有监督选取中心及正交最小二 乘法等。而最近邻聚类学习法是一种在线自适应聚类学习方法，不需要事先确 定隐藏层的节点数，完成聚类得到的RBF网络是最优的。该算法的具体步骤如 下：According to the different selection methods of RBF centers, the learning algorithms of RBF neural network can be roughly divided into random selection of centers, self-organization selection of centers, nearest neighbor clustering method, supervised selection of centers, and orthogonal least squares method. The nearest neighbor clustering learning method is an online adaptive clustering learning method, which does not need to determine the number of nodes in the hidden layer in advance, and the RBF network obtained by clustering is optimal. The specific steps of the algorithm are as follows:
①选取高斯函数宽度r，定义向量A(m)保存各类别输出向量之和，定义变量 B(m)用于保存各类的样本个数，m为类别数；① Select the Gaussian function width r, define the vector A(m) to save the sum of the output vectors of each category, define the variable B(m) to save the number of samples of each category, and m is the number of categories;
②从第一个数据对(x1，y1)开始，将该数据作为第一个聚类中心，令 c1＝x1，A(1)＝y1，B(1)＝1。该隐藏层单元到输出层的权值为②Starting from the first data pair (x1 , y1 ), this data is used as the first cluster center, and c1 =x1 , A(1)=y1 , B(1)=1. The weight of the hidden layer unit to the output layer is
③考虑第二个数据对(x2，y2),先求取x2到c1的距离即欧式范数|x2-c1|， 若|x2-c1|＜r，则x2属于第一个聚类，令A(1)＝y1+y2，B(1)＝2，若|x2-c1|＞r，则以x2为中心c2新建一个聚类，令A(1)＝y2，B(2)＝1，③Consider the second data pair (x2 , y2 ), first calculate the distance from x2 to c1 , which is the Euclidean norm |x2 -c1 |, if |x2 -c1 |<r, then x2 belongs to the first cluster, let A(1)=y1 +y2 , B(1)=2, If |x2 -c1 |＞r, create a new cluster with x2 as the center c2 , set A(1)=y2 , B(2)=1,
④考虑第k个数据对(xk，yk),假设之前已经构建了t个聚类(t＜k)，计算xk到 这t个聚类中心的距离di＝|xi-ci|(x＝1，2，...，t)。如果di的最小值小于r，则xk属于该中心定义的聚类，A(j)＝A(j)+yk，B(j)＝B(j)+1，如果 di＞r，则令ct+1＝xk，定义一个新的聚类。t＝t+1,A(j)＝+yk，B(j)＝1，④Consider the kth data pair (xk , yk ), assuming that t clusters have been constructed before (t<k), calculate the distance di = xi -c from xk to the center of the t clustersi |(x=1, 2, . . . , t). If the minimum value of di is less than r, then xk belongs to the cluster defined by this center, A(j)=A(j)+yk , B(j)=B(j)+1, If di >r, let ct+1 =xk to define a new cluster. t=t+1, A(j)=+yk , B(j)=1,
⑤对训练样本中每一个数据重复上述过程，即可构建RBF神经网络。最后 网络的输出可以表示为：⑤ Repeat the above process for each data in the training sample to construct the RBF neural network. The output of the final network can be expressed as:
3.3模型的使用与优化3.3 Model usage and optimization
完成RBF网络的构建后，将从新患者的三种信号中提取的特征与该患者的 基本生理指标及病史组成特征向量，将该特征向量输入筛查模型，即可得到一 个SYNTAX评分的估计值，并在显示屏上给出相对应的建议。After the construction of the RBF network is completed, the features extracted from the three signals of the new patient and the basic physiological indicators and medical history of the patient form a feature vector, and the feature vector is input into the screening model to obtain an estimated value of the SYNTAX score. And give corresponding suggestions on the display screen.
对于在筛查后又进行了冠脉造影的患者，可以将验证结果上传至远程服务 器，扩充现有的训练样本集，对筛查模型的参数进行进一步优化，不断提高该 模型对冠状动脉疾病预测的准确率。理论上，只要训练样本集足够大，则该模 型的效果将逼近现有的诊断“金标准”——冠脉造影。For patients who underwent coronary angiography after screening, the verification results can be uploaded to the remote server to expand the existing training sample set, further optimize the parameters of the screening model, and continuously improve the prediction of coronary artery disease by the model the accuracy rate. Theoretically, as long as the training sample set is large enough, the effect of the model will approach the existing "gold standard" of diagnosis - coronary angiography.
根据一个或多个实施例，一种信号特征提取方法，所述信号包括从用户 获取的心电信号、心音信号和脉搏波信号，对信号特征的提取方法包括以下步 骤：According to one or more embodiments, a signal feature extraction method, the signal includes an electrocardiogram signal, a heart sound signal and a pulse wave signal obtained from a user, and the signal feature extraction method includes the following steps:
对于心电信号，确定其ST段电平和QRS波群宽度；For the ECG signal, determine its ST segment level and QRS complex width;
对于心音信号，确定其频谱质心、频谱滚降点及通过小波变换得到的5个 频率区间的舒张期能量与整个心动周期能量的比值；For the heart sound signal, determine its spectral centroid, spectral roll-off point and the diastolic energy of 5 frequency intervals obtained by wavelet transform and the ratio of the whole cardiac cycle energy;
对于脉搏波信号，确定踝肱脉搏波传导速度(baPWV)和踝肱指数(ABI)。For the pulse wave signal, the ankle-brachial pulse wave velocity (baPWV) and the ankle-brachial index (ABI) were determined.
将心电信号、心音信号和脉搏波信号特征提取后，将特征向量作为训练样本 用于生成冠心病筛查模型。After extracting the features of the ECG signal, heart sound signal and pulse wave signal, the feature vector is used as a training sample to generate a coronary heart disease screening model.
进一步的，还可以将基本生理参数和病史类型参数作为冠心病筛查模型的训 练样本特征，例如，Further, the basic physiological parameters and medical history type parameters can also be used as the training sample features of the coronary heart disease screening model, for example,
基本生理参数包括：与血液检测指标有关的空腹血糖、甘油三酯、总胆固醇、 高密度脂蛋白胆固醇、低密度脂蛋白胆固醇、脂蛋白α，糖化血红蛋白以及用 户的年龄、性别或BMI指数的其中一种或任意组合；Basic physiological parameters include: fasting blood glucose, triglycerides, total cholesterol, high-density lipoprotein cholesterol, low-density lipoprotein cholesterol, lipoprotein α, glycated hemoglobin and the user's age, gender or BMI index related to blood test indicators. one or any combination;
病史类型参数包括：冠心病史、糖尿病史、吸烟史、饮酒史、高血脂症史、 PCI手术史或心脏旁路移植手术史的其中一种或任意组合。Medical history type parameters include: history of coronary heart disease, history of diabetes, history of smoking, history of drinking, history of hyperlipidemia, history of PCI operation or history of heart bypass transplantation, one or any combination.
值得说明的是，虽然前述内容已经参考若干具体实施方式描述了本发明创 造的精神和原理，但是应该理解，本发明并不限于所公开的具体实施方式，对 各方面的划分也不意味着这些方面中的特征不能组合，这种划分仅是为了表述 的方便。本发明旨在涵盖所附权利要求的精神和范围内所包括的各种修改和等 同布置。It is worth noting that although the foregoing content has described the spirit and principle of the invention with reference to several specific embodiments, it should be understood that the present invention is not limited to the disclosed specific embodiments, and the division of various aspects does not mean that these Features within an aspect cannot be combined, this division is for convenience of presentation only. The present invention is intended to cover various modifications and equivalent arrangements included within the spirit and scope of the appended claims.
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810286359.8ACN108577883A (en) | 2018-04-03 | 2018-04-03 | A kind of Screening for coronary artery disease device, screening system and signal characteristic extracting methods |
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810286359.8ACN108577883A (en) | 2018-04-03 | 2018-04-03 | A kind of Screening for coronary artery disease device, screening system and signal characteristic extracting methods |
| Publication Number | Publication Date |
|---|---|
| CN108577883Atrue CN108577883A (en) | 2018-09-28 |
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810286359.8APendingCN108577883A (en) | 2018-04-03 | 2018-04-03 | A kind of Screening for coronary artery disease device, screening system and signal characteristic extracting methods |
| Country | Link |
|---|---|
| CN (1) | CN108577883A (en) |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109272514A (en)* | 2018-10-05 | 2019-01-25 | 数坤（北京）网络科技有限公司 | The sample evaluation method and model training method of coronary artery parted pattern |
| CN109497992A (en)* | 2019-01-04 | 2019-03-22 | 济南汇医融工科技有限公司 | Coronary heart disease intelligence screening apparatus based on machine learning method |
| CN110353657A (en)* | 2019-07-16 | 2019-10-22 | 上海数创医疗科技有限公司 | It is a kind of based on double multiple waveforms type screening techniques and device for selecting mechanism |
| CN110353650A (en)* | 2019-07-18 | 2019-10-22 | 北京合众思壮时空物联科技有限公司 | Signature waveform extraction system based on PPG signal |
| CN110537910A (en)* | 2019-09-18 | 2019-12-06 | 山东大学 | Non-invasive screening system for coronary heart disease based on joint analysis of ECG and heart sound signals |
| CN111436921A (en)* | 2020-03-17 | 2020-07-24 | 深圳市贝斯曼精密仪器有限公司 | ABI detection device and blood vessel detection system |
| CN111685740A (en)* | 2020-06-08 | 2020-09-22 | 上海交通大学 | Heart function parameter detection method and device |
| CN111798980A (en)* | 2020-07-10 | 2020-10-20 | 哈尔滨工业大学（深圳） | Complex medical biological signal processing method and device based on deep learning network |
| CN112233750A (en)* | 2020-10-20 | 2021-01-15 | 吾征智能技术(北京)有限公司 | Information matching system based on hemoptysis character and disease |
| CN112308813A (en)* | 2019-07-26 | 2021-02-02 | 宏碁股份有限公司 | Blood vessel state evaluation method and blood vessel state evaluation device |
| CN112307804A (en)* | 2019-07-25 | 2021-02-02 | 宏碁股份有限公司 | Blood vessel state evaluation method and blood vessel state evaluation device |
| WO2021031979A1 (en)* | 2019-08-19 | 2021-02-25 | 华为技术有限公司 | Acquisition method and device for physiological parameter and processing method and device for physiological parameter |
| CN112472048A (en)* | 2021-01-15 | 2021-03-12 | 郑州大学 | Neural network structure for pulse condition recognition of cardiovascular disease patient |
| CN112494000A (en)* | 2020-11-26 | 2021-03-16 | 河北德睿健康科技有限公司 | Coronary artery stenosis visualization quantification method and equipment based on multiple paths of heart sounds |
| CN113053528A (en)* | 2021-04-23 | 2021-06-29 | 中国人民解放军北部战区总医院 | Coronary heart disease ischemia risk assessment method and system |
| CN113517046A (en)* | 2021-04-15 | 2021-10-19 | 中南大学 | Heterogeneous data feature fusion method in electronic medical record, prediction method and system based on fusion features and readable storage medium |
| CN113796889A (en)* | 2021-08-30 | 2021-12-17 | 西安交通大学 | Auxiliary electronic stethoscope signal discrimination method based on deep learning |
| CN114601435A (en)* | 2022-04-02 | 2022-06-10 | 北京雪扬科技有限公司 | A monitoring method for monitoring fetal movement |
| CN115336985A (en)* | 2021-05-13 | 2022-11-15 | 先导者股份有限公司 | Method and system for monitoring cardiac function based on heart sound centroid |
| CN115349834A (en)* | 2022-10-18 | 2022-11-18 | 合肥心之声健康科技有限公司 | Electrocardiogram screening method and system for asymptomatic severe coronary artery stenosis |
| CN115553784A (en)* | 2022-09-26 | 2023-01-03 | 山东大学 | Evaluation method and system for coronary heart disease based on coupled analysis of ECG and heart sound signals |
| CN115590469A (en)* | 2021-07-09 | 2023-01-13 | 特拉量子股份公司（Ch） | Method and system for converting physiological signals |
| CN115670902A (en)* | 2022-10-10 | 2023-02-03 | 山东大学 | Intelligent decision making system and method for cardio-pulmonary resuscitation |
| US11583194B1 (en) | 2020-03-20 | 2023-02-21 | Hemchandra Madhusudan Shertukde | Non-invasive angiography device |
| WO2024093723A1 (en)* | 2022-10-31 | 2024-05-10 | 歌尔科技有限公司 | Smartwatch and physiological data measurement method |
| CN118588263A (en)* | 2024-08-07 | 2024-09-03 | 吉林大学 | A hierarchical nursing management system for pediatric intensive care units based on early warning |
| US12357235B2 (en) | 2021-05-13 | 2025-07-15 | Pacesetter, Inc. | Method and system for monitoring heart function based on heart sound center of mass |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1265869A (en)* | 1999-03-04 | 2000-09-13 | 刘海祥 | Heart disease testing and positioning system |
| CN1371660A (en)* | 2001-02-22 | 2002-10-02 | 上海市脑血管病防治研究所 | Method for evaluating apoplexy dangerous posibility by cerebrovascular hemodynamics parameter |
| CN101176660A (en)* | 2007-12-06 | 2008-05-14 | 山东大学 | Method and device for detecting joint variability index of cardiovascular system |
| CN101530331A (en)* | 2008-09-11 | 2009-09-16 | 兰州理工大学 | Heart sound signal detection and analysis system based on Labview development platform |
| CN101645142A (en)* | 2008-08-04 | 2010-02-10 | 香港理工大学 | Fuzzy system for risk assessment of cardiovascular disease and stroke |
| CN102170821A (en)* | 2008-10-01 | 2011-08-31 | 株式会社Irumedi | Cardiovascular analysis system |
| CN103038765A (en)* | 2010-07-01 | 2013-04-10 | 诺基亚公司 | Method and apparatus for adapting a context model |
| CN104873186A (en)* | 2015-04-17 | 2015-09-02 | 中国科学院苏州生物医学工程技术研究所 | Wearable artery detection device and data processing method thereof |
| CN105208227A (en)* | 2014-06-24 | 2015-12-30 | 汤姆逊许可公司 | Method and system for setting detection parameters in an apparatus for on-hold music detection |
| CN105303054A (en)* | 2015-11-12 | 2016-02-03 | 清华大学 | Coronary heart disease condition assessment model and onset risk evaluation method based on linear time invariant system |
| US20170188862A1 (en)* | 2016-01-04 | 2017-07-06 | AventuSoft, LLC | System and Method of Extraction, Identification, Marking and Display of Heart Valve Signals |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1265869A (en)* | 1999-03-04 | 2000-09-13 | 刘海祥 | Heart disease testing and positioning system |
| CN1371660A (en)* | 2001-02-22 | 2002-10-02 | 上海市脑血管病防治研究所 | Method for evaluating apoplexy dangerous posibility by cerebrovascular hemodynamics parameter |
| CN101176660A (en)* | 2007-12-06 | 2008-05-14 | 山东大学 | Method and device for detecting joint variability index of cardiovascular system |
| CN101645142A (en)* | 2008-08-04 | 2010-02-10 | 香港理工大学 | Fuzzy system for risk assessment of cardiovascular disease and stroke |
| CN101530331A (en)* | 2008-09-11 | 2009-09-16 | 兰州理工大学 | Heart sound signal detection and analysis system based on Labview development platform |
| CN102170821A (en)* | 2008-10-01 | 2011-08-31 | 株式会社Irumedi | Cardiovascular analysis system |
| CN103038765A (en)* | 2010-07-01 | 2013-04-10 | 诺基亚公司 | Method and apparatus for adapting a context model |
| CN105208227A (en)* | 2014-06-24 | 2015-12-30 | 汤姆逊许可公司 | Method and system for setting detection parameters in an apparatus for on-hold music detection |
| CN104873186A (en)* | 2015-04-17 | 2015-09-02 | 中国科学院苏州生物医学工程技术研究所 | Wearable artery detection device and data processing method thereof |
| CN105303054A (en)* | 2015-11-12 | 2016-02-03 | 清华大学 | Coronary heart disease condition assessment model and onset risk evaluation method based on linear time invariant system |
| US20170188862A1 (en)* | 2016-01-04 | 2017-07-06 | AventuSoft, LLC | System and Method of Extraction, Identification, Marking and Display of Heart Valve Signals |
| Title |
|---|
| 张雪源: "面向音频检索的音频特征分析方法研究", 《中国博士学位论文全文数据库 信息科技辑》* |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109272514A (en)* | 2018-10-05 | 2019-01-25 | 数坤（北京）网络科技有限公司 | The sample evaluation method and model training method of coronary artery parted pattern |
| CN109497992A (en)* | 2019-01-04 | 2019-03-22 | 济南汇医融工科技有限公司 | Coronary heart disease intelligence screening apparatus based on machine learning method |
| CN110353657A (en)* | 2019-07-16 | 2019-10-22 | 上海数创医疗科技有限公司 | It is a kind of based on double multiple waveforms type screening techniques and device for selecting mechanism |
| CN110353650A (en)* | 2019-07-18 | 2019-10-22 | 北京合众思壮时空物联科技有限公司 | Signature waveform extraction system based on PPG signal |
| CN110353650B (en)* | 2019-07-18 | 2022-05-03 | 北京合众思壮时空物联科技有限公司 | PPG signal-based characteristic waveform extraction system |
| CN112307804A (en)* | 2019-07-25 | 2021-02-02 | 宏碁股份有限公司 | Blood vessel state evaluation method and blood vessel state evaluation device |
| CN112308813A (en)* | 2019-07-26 | 2021-02-02 | 宏碁股份有限公司 | Blood vessel state evaluation method and blood vessel state evaluation device |
| WO2021031979A1 (en)* | 2019-08-19 | 2021-02-25 | 华为技术有限公司 | Acquisition method and device for physiological parameter and processing method and device for physiological parameter |
| CN110537910B (en)* | 2019-09-18 | 2021-05-04 | 济南汇医融工科技有限公司 | Coronary heart disease noninvasive screening system based on electrocardio and heart sound signal joint analysis |
| CN110537910A (en)* | 2019-09-18 | 2019-12-06 | 山东大学 | Non-invasive screening system for coronary heart disease based on joint analysis of ECG and heart sound signals |
| CN111436921A (en)* | 2020-03-17 | 2020-07-24 | 深圳市贝斯曼精密仪器有限公司 | ABI detection device and blood vessel detection system |
| US11583194B1 (en) | 2020-03-20 | 2023-02-21 | Hemchandra Madhusudan Shertukde | Non-invasive angiography device |
| CN111685740B (en)* | 2020-06-08 | 2022-03-01 | 上海交通大学 | Method and device for detecting cardiac function parameters |
| CN111685740A (en)* | 2020-06-08 | 2020-09-22 | 上海交通大学 | Heart function parameter detection method and device |
| CN111798980A (en)* | 2020-07-10 | 2020-10-20 | 哈尔滨工业大学（深圳） | Complex medical biological signal processing method and device based on deep learning network |
| CN112233750B (en)* | 2020-10-20 | 2024-02-02 | 吾征智能技术(北京)有限公司 | Information matching system based on hemoptysis characters and diseases |
| CN112233750A (en)* | 2020-10-20 | 2021-01-15 | 吾征智能技术(北京)有限公司 | Information matching system based on hemoptysis character and disease |
| CN112494000A (en)* | 2020-11-26 | 2021-03-16 | 河北德睿健康科技有限公司 | Coronary artery stenosis visualization quantification method and equipment based on multiple paths of heart sounds |
| CN112472048A (en)* | 2021-01-15 | 2021-03-12 | 郑州大学 | Neural network structure for pulse condition recognition of cardiovascular disease patient |
| CN112472048B (en)* | 2021-01-15 | 2023-10-24 | 郑州大学 | Method for realizing neural network for identifying pulse condition of cardiovascular disease patient |
| CN113517046B (en)* | 2021-04-15 | 2023-11-07 | 中南大学 | Heterogeneous data feature fusion method in electronic medical record, fusion feature-based prediction method, fusion feature-based prediction system and readable storage medium |
| CN113517046A (en)* | 2021-04-15 | 2021-10-19 | 中南大学 | Heterogeneous data feature fusion method in electronic medical record, prediction method and system based on fusion features and readable storage medium |
| CN113053528A (en)* | 2021-04-23 | 2021-06-29 | 中国人民解放军北部战区总医院 | Coronary heart disease ischemia risk assessment method and system |
| CN115336985A (en)* | 2021-05-13 | 2022-11-15 | 先导者股份有限公司 | Method and system for monitoring cardiac function based on heart sound centroid |
| US12357235B2 (en) | 2021-05-13 | 2025-07-15 | Pacesetter, Inc. | Method and system for monitoring heart function based on heart sound center of mass |
| CN115336985B (en)* | 2021-05-13 | 2025-06-10 | 先导者股份有限公司 | Method and system for monitoring heart function based on heart sound centroid |
| CN115590469A (en)* | 2021-07-09 | 2023-01-13 | 特拉量子股份公司（Ch） | Method and system for converting physiological signals |
| CN113796889A (en)* | 2021-08-30 | 2021-12-17 | 西安交通大学 | Auxiliary electronic stethoscope signal discrimination method based on deep learning |
| CN114601435A (en)* | 2022-04-02 | 2022-06-10 | 北京雪扬科技有限公司 | A monitoring method for monitoring fetal movement |
| CN115553784B (en)* | 2022-09-26 | 2024-04-26 | 山东大学 | Coronary heart disease assessment method and system based on coupled analysis of electrocardiogram and heart sound signals |
| CN115553784A (en)* | 2022-09-26 | 2023-01-03 | 山东大学 | Evaluation method and system for coronary heart disease based on coupled analysis of ECG and heart sound signals |
| CN115670902A (en)* | 2022-10-10 | 2023-02-03 | 山东大学 | Intelligent decision making system and method for cardio-pulmonary resuscitation |
| CN115349834A (en)* | 2022-10-18 | 2022-11-18 | 合肥心之声健康科技有限公司 | Electrocardiogram screening method and system for asymptomatic severe coronary artery stenosis |
| WO2024093723A1 (en)* | 2022-10-31 | 2024-05-10 | 歌尔科技有限公司 | Smartwatch and physiological data measurement method |
| CN118588263A (en)* | 2024-08-07 | 2024-09-03 | 吉林大学 | A hierarchical nursing management system for pediatric intensive care units based on early warning |
| CN118588263B (en)* | 2024-08-07 | 2024-10-15 | 吉林大学 | Pediatric intensive care unit hierarchical nursing management system based on early warning |
| Publication | Publication Date | Title |
|---|---|---|
| CN108577883A (en) | A kind of Screening for coronary artery disease device, screening system and signal characteristic extracting methods | |
| US9220437B2 (en) | Diagnostic classifications of pulse signal waveform data | |
| Aydemir et al. | Classification of decompensated heart failure from clinical and home ballistocardiography | |
| Miao et al. | A wearable sensor for arterial stiffness monitoring based on machine learning algorithms | |
| Mai et al. | Non-contact heartbeat detection based on ballistocardiogram using UNet and bidirectional long short-term memory | |
| RU2657384C2 (en) | Method and system for noninvasive screening physiological parameters and pathology | |
| Zhang et al. | Cardiac arrhythmia classification by time–frequency features inputted to the designed convolutional neural networks | |
| CN113080991A (en) | Method, system, diagnosis device and storage medium for predicting and diagnosing heart failure based on CNN model and LSTM model | |
| CN111419205A (en) | A three-element cuffless continuous blood pressure detection system based on artificial neural network | |
| Al Younis et al. | Non-invasive technologies for heart failure, systolic and diastolic dysfunction modeling: a scoping review | |
| Argha et al. | Blood pressure estimation from korotkoff sound signals using an end-to-end deep-learning-based algorithm | |
| CN113598724A (en) | Device for assessing cardiovascular autonomic function | |
| Akouz et al. | A comprehensive review on monitoring sensors for cardiovascular disease prevention and management | |
| Puspasari et al. | A new heart sound signal identification approach suitable for smart healthcare systems | |
| Chen et al. | Beat-to-beat heart rate detection based on seismocardiogram using BiLSTM network | |
| Zia et al. | Harnessing the manifold structure of cardiomechanical signals for physiological monitoring during hemorrhage | |
| Pu et al. | Cuff-less blood pressure estimation from electrocardiogram and photoplethysmography based on VGG19-LSTM network | |
| Hu et al. | Feature extraction and choice in PCG based on Hilbert Transfer | |
| Sadek et al. | Detecting Cardiovascular Disease From PPG Signals using Machine Learning | |
| Lin et al. | Korotkoff sounds dynamically reflect changes in cardiac function based on deep learning methods | |
| Tang et al. | Data-knowledge co-driven feature based prediction model via photoplethysmography for evaluating blood pressure | |
| Mathe et al. | Advancements in Noise Reduction Techniques in ECG Signals: A Review | |
| Sahay et al. | Computer‐Aided Interpretation of ECG Signal—A Challenge | |
| Yuwono et al. | Design of smart electrocardiography (ECG) using Modified K-Nearest Neighbor (MKNN) | |
| Sooriamoorthy et al. | Emerging Technologies and Concepts for Cardiovascular Risk Detection |
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication | Application publication date:20180928 |