WO2023082402A1 - 一种基于机器学习的非侵入性心脏病诊断方法和装置 - Google Patents

一种基于机器学习的非侵入性心脏病诊断方法和装置 Download PDF

Info

Publication number
WO2023082402A1
WO2023082402A1 PCT/CN2021/137920 CN2021137920W WO2023082402A1 WO 2023082402 A1 WO2023082402 A1 WO 2023082402A1 CN 2021137920 W CN2021137920 W CN 2021137920W WO 2023082402 A1 WO2023082402 A1 WO 2023082402A1
Authority
WO
WIPO (PCT)
Prior art keywords
feature vector
signal
features
sample space
ecg
Prior art date
Application number
PCT/CN2021/137920
Other languages
English (en)
French (fr)
Inventor
黄建龙
吴剑煌
谢炜芳
Original Assignee
中国科学院深圳先进技术研究院
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 中国科学院深圳先进技术研究院 filed Critical 中国科学院深圳先进技术研究院
Publication of WO2023082402A1 publication Critical patent/WO2023082402A1/zh

Links

Images

Classifications

    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
    • A61B5/316—Modalities, i.e. specific diagnostic methods
    • A61B5/318—Heart-related electrical modalities, e.g. electrocardiography [ECG]
    • A61B5/346—Analysis of electrocardiograms
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/02—Detecting, measuring or recording pulse, heart rate, blood pressure or blood flow; Combined pulse/heart-rate/blood pressure determination; Evaluating a cardiovascular condition not otherwise provided for, e.g. using combinations of techniques provided for in this group with electrocardiography or electroauscultation; Heart catheters for measuring blood pressure
    • A61B5/024—Detecting, measuring or recording pulse rate or heart rate
    • A61B5/02416—Detecting, measuring or recording pulse rate or heart rate using photoplethysmograph signals, e.g. generated by infrared radiation
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
    • A61B5/316—Modalities, i.e. specific diagnostic methods
    • A61B5/318—Heart-related electrical modalities, e.g. electrocardiography [ECG]
    • A61B5/346—Analysis of electrocardiograms
    • A61B5/349—Detecting specific parameters of the electrocardiograph cycle
    • A61B5/352—Detecting R peaks, e.g. for synchronising diagnostic apparatus; Estimating R-R interval
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
    • A61B5/316—Modalities, i.e. specific diagnostic methods
    • A61B5/318—Heart-related electrical modalities, e.g. electrocardiography [ECG]
    • A61B5/346—Analysis of electrocardiograms
    • A61B5/349—Detecting specific parameters of the electrocardiograph cycle
    • A61B5/366—Detecting abnormal QRS complex, e.g. widening
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7235—Details of waveform analysis
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7235—Details of waveform analysis
    • A61B5/7253—Details of waveform analysis characterised by using transforms
    • A61B5/726—Details of waveform analysis characterised by using transforms using Wavelet transforms
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7235—Details of waveform analysis
    • A61B5/7264—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7235—Details of waveform analysis
    • A61B5/7264—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems
    • A61B5/7267—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems involving training the classification device

Definitions

  • the present invention relates to the technical field of medical image analysis, and more specifically, to a non-invasive heart disease diagnosis method and device based on machine learning.
  • ECG Electrocardiogram
  • PPG Photoplethysmography
  • the purpose of the present invention is to overcome the defects of the above-mentioned prior art, and provide a non-invasive heart disease diagnosis method and device based on machine learning.
  • a non-invasive heart disease diagnosis method based on machine learning includes the following steps:
  • the sample space of the first feature vector and the sample space of the second feature vector are fused, and the fused features are input to a pre-trained classifier to obtain a heart diagnosis result.
  • a non-invasive heart disease diagnosis device based on machine learning includes:
  • the first data collection unit used to collect the ECG signal of the target and detect the corresponding QRS complex
  • the second data collection unit used to collect the pulse wave signal of the target
  • the first feature extraction unit for extracting ECG signal features based on the QRS wave group, constructing a sample space for the first feature vector;
  • the second feature extraction unit used to extract the features of the pulse wave signal, constructing the obtained feature vector matrix as the sample space of the second feature vector;
  • Classification identification unit used to fuse the sample space of the first feature vector and the sample space of the second feature vector, and input the fusion features into a pre-trained classifier to obtain a heart diagnosis result.
  • the present invention has the advantage that it proposes a novel heart recognition method based on the combination of ECG and PPG, and puts the features of the PPG signal and the ECG signal into a classifier for recognition by combining the features of the PPG signal and the feature of the ECG signal.
  • the accuracy of heart recognition by machine learning methods is a novel heart recognition method based on the combination of ECG and PPG, and puts the features of the PPG signal and the ECG signal into a classifier for recognition by combining the features of the PPG signal and the feature of the ECG signal.
  • Fig. 1 is a flowchart of a non-invasive heart disease diagnosis method based on machine learning according to an embodiment of the present invention
  • Fig. 2 is a schematic diagram of an electrocardiographic cycle according to an embodiment of the present invention.
  • Fig. 3 is a structural comparison diagram of a recurrent neural network and a long-short-term memory network according to an embodiment of the present invention
  • FIG. 4 is a schematic structural diagram of a long short-term memory network according to an embodiment of the present invention.
  • Fig. 5 is a flow chart of fusion recognition of ECG and PPG feature layers based on multi-set discriminant correlation analysis according to an embodiment of the present invention.
  • the present invention uses computer-aided diagnosis of heart disease, and carries out relevant research on the classification of ECG and PPG through machine learning related knowledge. For example, firstly, use LSTM to extract the features of the PPG signal; then, according to the results of feature extraction, the sample is cut; use the Pan-Tompkins algorithm, use the collected ECG data set to perform feature extraction for the QRS complex of the ECG signal, and perform Cluster analysis of samples to eliminate some special cut-off samples; after the data screening is completed, sample features are selected, and principal component analysis (PCA) is used to reduce dimensionality; finally, sample features are used to train Classifier, and then use the trained classifier to achieve the cardiac diagnosis result of the target patient.
  • PCA principal component analysis
  • the provided non-invasive heart disease diagnosis method based on machine learning includes the following steps.
  • Step S110 extracting QRS complex information in the electrocardiogram signal.
  • the ECG signal records the voltage generated by the depolarization of different parts of the heart over a period of time.
  • the QRS complex reflects the rapid depolarization of the left and right ventricles.
  • Electrocardiographic signals are periodic. Compared with other electrical signals such as brain waves and muscle waves, they belong to a type of electrical waves that are easier to judge and distinguish.
  • the necessary link is to extract its QRS wave group, and the analysis of the ECG signal based on the QRS wave group is beneficial to improve the robustness and stability.
  • the QRS wave group is the potential change of ventricular depolarization. As shown in Figure 2, in the electrocardiogram signal, after the P wave ends, it enters the PR interval, and then the Q wave will show a downward trend, and the R wave will quickly Up, S wave followed by down, and the connected electric waves corresponding to these three waves are collectively referred to as QRS complex in this paper.
  • the R wave is the narrowest and largest wave of all waveforms.
  • the T wave corresponds to the repolarization process of the human ventricular muscle.
  • the process of depolarization of the heart corresponds to the P wave, so it is a blunt circular image.
  • PR corresponds to the conduction process. In an ideal state, there is actually no potential change, that is, the ECG signal is flat.
  • the QT period corresponds to a process of human heart ventricular contraction, so the change of the QT interval can effectively judge the human heart rate, and the size of the QT period is directly proportional to the speed of the human heart rate.
  • the methods for extracting and detecting QRS complexes include wavelet transform, neural network, Pan-Tompkins algorithm and its improved algorithm, improved wavelet algorithm with notch filter added, etc.
  • the wavelet transform method has high accuracy, but the disadvantage is that the calculation steps are very cumbersome, so the calculation efficiency is low, and the data redundancy will increase during the wavelet transform process.
  • the neural network can accurately detect the QRS complex to a certain extent, but the process of training the neural network is very long, and a large number of samples need to be collected. In addition, the training results produced by different training samples are also different. The signal difference is large, and the trained neural network model is not suitable for practical applications.
  • the Pan-Tompkins algorithm is used to detect the QRS complex information in the electrocardiogram. Compared with other QRS complex extraction algorithms, the Pan-Tompkins algorithm has high computational efficiency and does not require model training.
  • using the Pan-Tompkins method to detect R-wave peaks includes: filtering with a filter; calculating the first derivative of the filtered signal; squaring the signal after derivation; integrating the signal through a sliding window; using The threshold method detects processes such as the peak value of the R wave after processing.
  • the features of the ECG signal extraction include:
  • Step S111 using a filter to perform noise reduction processing.
  • noise processing in ECG signal data is the core content of data preprocessing.
  • a high-pass filter and a low-pass filter are used to filter out noise in the signal, thereby improving the signal-to-noise ratio of the electrocardiographic signal.
  • the main noises in ECG signals are power frequency interference and baseline drift.
  • the interference signal brought by the power system is called power frequency interference, and the frequency is generally concentrated at 50Hz. Skin contact, human breathing, etc. will produce the so-called baseline drift, and the frequency is generally concentrated in the range of 0.05Hz to 2Hz.
  • the frequency of the QRS complex is mainly concentrated in the range of 5Hz to 11Hz.
  • the cutoff frequency of the low-pass filter is 11Hz, and the difference equation can be expressed as:
  • the cut-off frequency of the high-pass filter is 2Hz, and the difference equation can be expressed as:
  • Step S112 performing amplification processing to detect R waves.
  • the R wave in the QRS complex is the steepest wave in an ECG cycle.
  • the specific difference equation is as follows:
  • Step S113 determine other features outside the R wave
  • a moving window integration is taken to determine other relevant features.
  • the purpose is to extract other features of the wave group. Assuming that the number of samples in the integration window is N, the difference equation is expressed as:
  • Step S114 threshold dynamic adjustment and search
  • Threshold dynamic adjustment and search belong to the core part of the algorithm, through which the peak value of ECG signal is searched, the threshold is updated through a certain calculation formula, and combined with backtracking detection, two-way discrimination and other content to detect. This process can be realized by using the existing technology, and will not be repeated here.
  • Step S120 using the long-short-term memory network to extract pulse wave features.
  • the long-short-term memory network does not have the vanishing gradient problem, and can memorize the input of thousands of discrete time steps before. Compared with the traditional RNN, the LSTM can perform better in longer sequences. LSTM replaces the activation function of neurons with a unit with a "gate" control structure, and each unit is called a cell (Cell).
  • the present invention uses LSTM to use a storage unit with a forgetting gate.
  • LSTM has two transfer states c(t) (cell state, cell state) and h(t) (hidden state, hidden state), h(t) in RNN is equivalent to c(t) in LSTM.
  • LSTM is very slow to change the c(t) passed on.
  • the output c(t) is the c(t-1) passed from the previous state plus some values, and h(t) often has a lot of different values under different nodes. The big difference is that these highly fluctuating h(t) produce the effect of forgetting and selecting memory.
  • W represents the weight of the corresponding item.
  • the characteristic parameters conforming to the physiological law are obtained through forgetting, selective memory, output stage and the like.
  • the PPG signal contains a large number of features related to heart rate, which are beneficial to the LSTM network to improve the prediction accuracy.
  • GRU Gate Recurrent Unit
  • Step S130 fusing the ECG signal features and pulse wave features extracted based on the QRS complex.
  • K-Means and other clustering can be used for data screening after truncating samples for further feature fusion.
  • the feature fusion process includes:
  • Step S131 the feature vector obtained after feature extraction of the ECG signal forms the sample space X ⁇ Rp ⁇ n of the first feature.
  • ECG signal characteristics reflect wave group shape characteristics and interval characteristics, etc.
  • the features of the ECG signal include time-domain features of the ECG signal, transform domain features of the ECG signal, and the like.
  • Step S132 the feature vector matrix obtained after feature extraction of the PPG signal forms a second feature sample space Y ⁇ Rq ⁇ n.
  • Step S133 using the maximum and minimum value method to perform a normalization operation on the eigenvectors, so that the order of magnitude of the two groups of eigenvectors is close.
  • normalization processing is performed on the feature vectors to eliminate adverse effects of singular sample data on subsequent prediction results.
  • Step S140 using fusion features to obtain heart disease classification results.
  • a classifier is trained using the fusion features, and then the trained classifier is used for real-time heart disease diagnosis analysis.
  • a Support Vector Machine is chosen as the classifier. This is because SVM is the classifier with the largest interval defined in the feature space, and is more suitable for small sample learning, and the final decision result of SVM depends on a small number of support vectors, so that it can "eliminate” a large number of redundant samples, and the calculation efficiency is relatively high. High and has better "robustness”.
  • ECG samples and PPG signal samples of a large number of heart disease patients may not be easy to obtain, and in practical applications, the relevant data of patients may vary greatly, so SVM can be used to obtain more accurate diagnostic results, such as normal heartbeat, left ventricular artery obstruction Heart rate, right ventricular artery obstruction heart rate, atrial fibrillation, ventricular systole and atrial systole, etc.
  • principal component analysis can also be used to select key features as the input of the classifier.
  • the present invention also provides a non-invasive heart disease diagnosis device based on machine learning, which is used to realize one or more aspects of the above method.
  • the device includes: a first data collection unit, which is used to collect the ECG signal of the target and detect the corresponding QRS complex; a second data collection unit, which is used to collect the pulse wave signal of the target; a first feature extraction unit , which is used to extract the features of the electrocardiogram signal based on the QRS wave group, and constructs the sample space of the first feature vector; the second feature extraction unit, which is used to extract the features of the pulse wave signal, constructs the obtained feature vector matrix is the sample space of the second feature vector; a classification recognition unit, which is used to fuse the sample space of the first feature vector and the sample space of the second feature vector, and input the fusion feature to the pre-trained classifier, Obtain cardiac diagnosis results.
  • Each unit involved in the device can be implemented by using a processor, FPGA or special hardware.
  • the present invention uses the data of patients in the hospital to collect the pulse wave signal and ECG of the patient respectively, and first uses LSTM to extract the features of the PPG signal. Secondly, using the Pan-Tompkins algorithm, the collected ECG data set is used to extract the features of the QRS complex of the ECG signal, and the K-Means clustering is used for data screening after cutting the sample. Secondly, the extracted PPG signal features and ECG signal features are fused, and finally the fused features are classified by a classifier.
  • the present invention proposes a novel heart recognition method based on the combination of ECG and PPG, which combines ECG and PPG for the first time, provides a supplementary solution for heart recognition, and improves the accuracy of heart recognition using machine learning, and obtains Cardiac diagnostic classification results can be used for clinical indication.
  • the present invention can be a system, method and/or computer program product.
  • a computer program product may include a computer readable storage medium having computer readable program instructions thereon for causing a processor to implement various aspects of the present invention.
  • a computer readable storage medium may be a tangible device that can retain and store instructions for use by an instruction execution device.
  • a computer readable storage medium may be, for example, but is not limited to, an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the foregoing.
  • Computer-readable storage media include: portable computer diskettes, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), or flash memory), static random access memory (SRAM), compact disc read only memory (CD-ROM), digital versatile disc (DVD), memory stick, floppy disk, mechanically encoded device, such as a printer with instructions stored thereon A hole card or a raised structure in a groove, and any suitable combination of the above.
  • RAM random access memory
  • ROM read-only memory
  • EPROM erasable programmable read-only memory
  • flash memory static random access memory
  • SRAM static random access memory
  • CD-ROM compact disc read only memory
  • DVD digital versatile disc
  • memory stick floppy disk
  • mechanically encoded device such as a printer with instructions stored thereon
  • a hole card or a raised structure in a groove and any suitable combination of the above.
  • computer-readable storage media are not to be construed as transient signals per se, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagating through waveguides or other transmission media (e.g., pulses of light through fiber optic cables), or transmitted electrical signals.
  • Computer-readable program instructions described herein may be downloaded from a computer-readable storage medium to a respective computing/processing device, or downloaded to an external computer or external storage device over a network, such as the Internet, a local area network, a wide area network, and/or a wireless network.
  • the network may include copper transmission cables, fiber optic transmission, wireless transmission, routers, firewalls, switches, gateway computers, and/or edge servers.
  • a network adapter card or a network interface in each computing/processing device receives computer-readable program instructions from the network and forwards the computer-readable program instructions for storage in a computer-readable storage medium in each computing/processing device .
  • Computer program instructions for carrying out operations of the present invention may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-related instructions, microcode, firmware instructions, state setting data, or Source or object code written in any combination, including object-oriented programming languages—such as Smalltalk, C++, Python, etc., and conventional procedural programming languages—such as the “C” language or similar programming languages.
  • Computer readable program instructions may execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer, or entirely on the remote computer or server implement.
  • the remote computer can be connected to the user computer through any kind of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computer (such as via the Internet using an Internet service provider). connect).
  • LAN local area network
  • WAN wide area network
  • an electronic circuit such as a programmable logic circuit, field programmable gate array (FPGA), or programmable logic array (PLA)
  • FPGA field programmable gate array
  • PDA programmable logic array
  • These computer-readable program instructions may be provided to a processor of a general purpose computer, special purpose computer, or other programmable data processing apparatus to produce a machine such that when executed by the processor of the computer or other programmable data processing apparatus , producing an apparatus for realizing the functions/actions specified in one or more blocks in the flowchart and/or block diagram.
  • These computer-readable program instructions can also be stored in a computer-readable storage medium, and these instructions cause computers, programmable data processing devices and/or other devices to work in a specific way, so that the computer-readable medium storing instructions includes An article of manufacture comprising instructions for implementing various aspects of the functions/acts specified in one or more blocks in flowcharts and/or block diagrams.
  • each block in a flowchart or block diagram may represent a module, a portion of a program segment, or an instruction that includes one or more Executable instructions.
  • the functions noted in the block may occur out of the order noted in the figures. For example, two blocks in succession may, in fact, be executed substantially concurrently, or they may sometimes be executed in the reverse order, depending upon the functionality involved.
  • each block of the block diagrams and/or flowchart illustrations, and combinations of blocks in the block diagrams and/or flowchart illustrations can be implemented by a dedicated hardware-based system that performs the specified function or action , or may be implemented by a combination of dedicated hardware and computer instructions. It is well known to those skilled in the art that implementation by means of hardware, implementation by means of software, and implementation by a combination of software and hardware are all equivalent.

Abstract

本发明公开了一种基于机器学习的非侵入性心脏病诊断方法和装置。该方法包括：采集目标的心电图信号并检测对应的QRS波群；采集该目标的脉搏波信号；基于所述QRS波群提取心电信号特征，构建为第一特征向量的样本空间；提取所述脉搏波信号的特征，将得到的特征向量矩阵构建为第二特征向量的样本空间；将所述第一特征向量的样本空间和所述第二特征向量的样本空间进行融合，利用融合特征输入到预训练的分类器，获得心脏诊断结果。本发明提高了心脏识别的准确率和效率。

Description

一种基于机器学习的非侵入性心脏病诊断方法和装置 技术领域
本发明涉及医学图像分析技术领域，更具体地，涉及一种基于机器学习的非侵入性心脏病诊断方法和装置。
背景技术
随着心脏病患者的不断增加，有限的医疗资源无法满足日益增长的心血管疾病患者的就医需求。目前，心脏病已经成为危害人类健康的重大疾病。近年来，人工智能在医疗行业中逐渐受到人们关注，大量机器学习算法在医疗保健领域广泛应用，涉及智能风险评估、辅助诊断、医学影像、药物研发等多个领域，由此可见“互联网+医疗健康”模式将在未来成为人们日常生活的一部分。心电图(ECG)是医生诊断心血管疾病的重要依据，能真实地反映心脏的健康状况。同时，随着传感器技术的进步，基于脉搏波信号(PPG)的可穿戴设备越来越多，人们可以通过这些可穿戴设备收集人体生理体征信号，随时监控自身健康状况。
在传统的心脏诊断中，医生通过查看心电图(Electrocardiography，ECG)判断心脏病，但这种诊断方式带有主观性，且耗时耗力。Kora等基于ECG使用Cuckoo Search算法实现自动识别心脏病；王凯基于ECG的自动心律失常检测模型能够有效监测心脏异常信号的ECG信号特征。心脏病监测的困难主要在于阵发性房颤，此类心脏病的特点是突发突止、持续时间短，24h动态心电图常作为首选的监测方法。李芳等通过分析总结心脏病的心电散点图特征，分析24h动态心电图来诊断心脏病。武求花等运用Lorenz-RR散点图(Lorenz散点图)与时间-RR间期散点图(时间散点图)及逆向技术相结合的方法，快速识别不同发作时间的PAF，为临床提供更快更准确的心电诊断技术。目前国内外对于心脏识别的研究大多基于ECG，但ECG设备普遍有操作复杂、携带困难、价格较高、多通道等局限性。光 电容积脉搏波描记(PPG)是使用激光对待测流体或固体的体积进行检测与记录的方法，可以用来监测毛细血管中血液的体积变化。有研究表明。PPG方法采集信息容易、方便携带，相较于ECG方法，PPG更加擅长长期监测。但是PPG可能没有ECG准确。
综上，在现有技术中，通常仅采用ECG或PPG进行心脏诊断分析，其准确性和效率还有待改进。
发明内容
本发明的目的是克服上述现有技术的缺陷，提供一种基于机器学习的非侵入性心脏病诊断方法和装置。
根据本发明的第一方面，提供一种基于机器学习的非侵入性心脏病诊断方法。该方法包括以下步骤：
采集目标的心电图信号并检测对应的QRS波群；
采集该目标的脉搏波信号；
基于所述QRS波群提取心电图信号特征，构建为第一特征向量的样本空间；
提取所述脉搏波信号的特征，将得到的特征向量矩阵构建为第二特征向量的样本空间；
将所述第一特征向量的样本空间和所述第二特征向量的样本空间进行融合，利用融合特征输入到预训练的分类器，获得心脏诊断结果。
根据本发明的第二方面，提供一种基于机器学习的非侵入性心脏病诊断装置。该装置包括：
第一数据采集单元：用于采集目标的心电图信号并检测对应的QRS波群；
第二数据采集单元：用于采集该目标的脉搏波信号；
第一特征提取单元：用于基于所述QRS波群提取心电图信号特征，构建为第一特征向量的样本空间；
第二特征提取单元：用于提取所述脉搏波信号的特征，将得到的特征向量矩阵构建为第二特征向量的样本空间；
分类识别单元：用于将所述第一特征向量的样本空间和所述第二特征向量的样本空间进行融合，利用融合特征输入到预训练的分类器，获得心脏诊断结果。
与现有技术相比，本发明的优点在于，提出了基于ECG和PPG相结合的新型心脏识别方法，通过将PPG信号的特征和ECG信号的特征融合，放入分类器进行识别，提高了利用机器学习的方法对心脏识别的准确率。
通过以下参照附图对本发明的示例性实施例的详细描述，本发明的其它特征及其优点将会变得清楚。
附图说明
被结合在说明书中并构成说明书的一部分的附图示出了本发明的实施例，并且连同其说明一起用于解释本发明的原理。
图1是根据本发明一个实施例的基于机器学习的非侵入性心脏病诊断方法的流程图；
图2是根据本发明一个实施例的心电周期示意图；
图3是根据本发明一个实施例的循环神经网络与长短时记忆网络的结构对比图；
图4是根据本发明一个实施例的长短时记忆网络的结构示意图；
图5是根据本发明一个实施例的基于多集判别相关分析的ECG和PPG特征层融合识别流程图。
具体实施方式
现在将参照附图来详细描述本发明的各种示例性实施例。应注意到：除非另外具体说明，否则在这些实施例中阐述的部件和步骤的相对布置、数字表达式和数值不限制本发明的范围。
以下对至少一个示例性实施例的描述实际上仅仅是说明性的，决不作为对本发明及其应用或使用的任何限制。
对于相关领域普通技术人员已知的技术、方法和设备可能不作详细讨论，但在适当情况下，所述技术、方法和设备应当被视为说明书的一部分。
在这里示出和讨论的所有例子中，任何具体值应被解释为仅仅是示例性的，而不是作为限制。因此，示例性实施例的其它例子可以具有不同的值。
应注意到：相似的标号和字母在下面的附图中表示类似项，因此，一旦某一项在一个附图中被定义，则在随后的附图中不需要对其进行进一步讨论。
本发明利用计算机辅助诊断心脏病，通过机器学习相关的知识分别对于ECG和PPG的分类进行了相关的研究。例如，首先，利用LSTM提取PPG信号的特征；然后，根据特征提取的结果进行了样本的截裁；利用Pan-Tompkins算法，使用收集的心电图数据集对于心电信号QRS波群进行特征提取，进行样本的聚类分析，以剔除了部分特殊的截裁小样；在数据筛选完成后，进行样本特征的选择，利用主成分分析法(PCA)，进行了降低维度的处理；最后，利用样本特征训练分类器，进而利用经训练的分类器实现对目标患者的心脏诊断结果。
具体地，结合图1和图5所示，所提供的基于机器学习的非侵入性心脏病诊断方法包括以下步骤。
步骤S110，提取心电图信号中QRS波群信息。
心电图信号记录一定时间内心脏不同部位去极化产生的电压。QRS波群反映了左右心室的快速去极化的过程。心电信号具有周期性，相对于其它电信号例如脑电波，肌肉电波，属于比较容易判断分辨特征的一类电波。分析心电信号，必要环节是提取其QRS波群，基于QRS波群进行心电信号分析，有利于提高鲁棒性和稳定性。
QRS波群是心室去极化的电位变化，如图2所示，在心电图信号中，在P波结束后，进入PR区间，之后Q波会呈现出一个向下的趋势，R波则会快速向上，S波紧接着向下，这三个波所对应的连接电波，本文统称为QRS波群。
由图2可以明显看出，R波是所有波形中最窄，幅度最大的波。T波对应的是人体心室肌肉的复极过程。心脏去极化的过程对应P波，因此它是钝圆形的图像。PR对应的是传导过程，在理想状态下实际上是没有电位 变化的，即心电信号为平。QT期间对应人体心脏心室收缩的一个过程，因此通过QT区间的变化可以有效的判断人的心率，QT期间的大小与人心率的快慢呈正比关系。
在正常的心电信号的周期中，P波和T波没有明显的特征，为了明显分辨心电信号，基本上还是需要通过QRS波群的位置来对其进行判断。选择QRS波群还有一个非常重要的原因是因为QRS波群是心电信号中最明显的信号，正确选择了QRS波群才能进一步完成对其它信号波的选择。
对于QRS波群的提取和检测方法包括小波变换、神经网络和Pan-Tompkins算法及其改良算法、增加了陷波器的小波改良算法等。
小波变换方法准确性高，但是缺点在于计算步骤非常繁琐，因此计算效率较低，并且在小波变换过程中会导致数据的冗余度增加。
神经网络一定程度上可以准确检测QRS波群，但训练神经网络的过程非常漫长的，并且需要采集大量的样本，此外不同的训练样本产生的训练结果也并不相同，而由于心脏病患者的心电图信号差异较大，经训练的神经网络模型并不适用于实际的应用。
为了平衡心电信号分析的准确性和计算效率，在一个实施例中，采用Pan-Tompkins算法检测心电图中QRS波群信息。相对于其他的QRS波群提取算法，Pan-Tompkins算法计算效率高并且不需要模型训练。简言之，利用Pan-Tompkins法检测R波峰值包括：利用滤波器进行滤波；对滤波后的信号求一阶导数；对求导之后的信号进行平方运算；将信号通过滑动窗口进行积分；使用阈值法检测经过处理之后的R波峰值等过程。
具体地，基于Pan-Tompkins算法提取心电图信号的特征包括：
步骤S111，利用滤波器进行降噪处理。
因为信号采集设备和环境的不同，在心电信号采集过程中会引入不同的噪声，所以对于心电信号数据中的噪声处理是数据预处理的核心内容。
在一个实施例中，采用高通滤波器和低通滤波器来滤除信号中的噪声，从而提升心电信号的信噪比。
心电信号中主要的噪声为工频干扰和基线漂移。电力系统等带入的干扰信号称为工频干扰，频率一般集中在50Hz。而皮肤的接触，人的呼吸等 都会产生所谓的基线漂移，频率一般集中在0.05Hz至2Hz。而QRS波群的频率主要集中在5Hz至11Hz。
以T为采样周期，低通滤波器的截止频率为11Hz，差分方程可表示为：
y(nT)＝2y(nT-T)-y(nT-2T)
-2x(nT-6T)+x(nT-12T)+x(nT)    (1)
高通滤波器的截至频率为2Hz，差分方程可表示为：
y(nT)＝y(nT-T)-x(nT)
+32x(nT-16T)+x(nT-32T)   (2)
步骤S112，进行放大处理以检测出R波。
QRS波群中R波是位于一个心电信号周期中最陡的一个波。为了方便将QRS波从其他的心电信号波中区分出来，需要利用相应的微分方法对于QRS波群进行陡坡特性的放大，具体的差分方程如下：
y(nT)＝x(nT)-x(xT-2T)   (3)
高频特性可以通过非线性平方函数来增强，其方程可以表示为：
y(nT)＝[x(nT)] 2   (4)
步骤S113，确定R波外的其他特征
上述的方法基本都是围绕R波来展开进行的，但单纯的对于R波的处理对于心电信号的异常判断无疑是不够完整的。因此对于波群的其它特征同样也需要进行确定。
在一个实施例中，采取了一个移动窗积分来确定其它相关的特征。目的是为了提取波群的其他特征。设积分窗内样本数为N，差分方程表示为：
Figure PCTCN2021137920-appb-000001
步骤S114，阈值动态调整和搜索
阈值动态调整和搜索属于该算法的核心部分，通过其搜索心电信号峰值，通过一定的计算式更新阈值，并结合回溯检测，双向判别等内容来检测。该过程可采用现有技术实现，在此不再赘述。
步骤S120，利用长短时记忆网络提取脉搏波特征。
与传统RNN不同，长短时记忆网络(LSTM)不存在消失梯度问题， 可记忆之前数千个离散时间步长的输入，与传统RNN相比，LSTM可在更长的序列中有更好表现。LSTM将神经元的激活功能替换为一个具有“门”控制结构的单元，每个单元称为一个细胞(Cell)，本发明采用LSTM使用具有遗忘门的存储单元。
如图3所示，相比于RNN只有一个传递状态h(t)，LSTM则有两个传输状态c(t)(cell state，细胞状态)和h(t)(hidden state，隐藏状态)，RNN中的h(t)相当于LSTM的c(t)。LSTM对于传递下去的c(t)改变很慢，通常输出的c(t)是上一个状态传过来的c(t-1)加上一些数值，而h(t)在不同节点下往往有很大区别，正是这些波动较大的h(t)产生忘记和选择记忆的效果。
首先使用当前输入x(t)和上一个状态传递下来的h(t-1)拼接训练得到4中状态(如图4所示)：z、z i、z f、z o，其中z f、z i、z o分别作为遗忘门(forget gate)，输入门(input gate)和输出门(output gate)的控制信号，z是神经网络输入层的输入。4种状态计算公式分别为：
z i(t)＝σ(W xi⊙x(t)+W hi⊙h(t-1))  (6)
Figure PCTCN2021137920-appb-000002
z o(t)＝σ(W xo⊙x(t)+W ho⊙h(t-1))   (8)
z f(t)＝σ(W xf⊙x(t)+W hf⊙h(t-1))   (9)
其中，W表示对应项的权重。
在该步骤中，对于光电容积脉搏波数据(PPG)序列通过忘记、选择记忆、输出阶段等得出符合生理规律的特征参数。PPG信号中包含大量例如与心率相关的特征，这些特征有利于LSTM网络提高预测精度。需说明的是，也可采用GRU(门控循环单元)替代LSTM。
步骤S130，融合基于QRS波群提取的心电信号特征和脉搏波特征。
在获得ECG信号特征和PPG信号特征后，可以在截裁样本后使用K-Means等聚类进行数据筛选，以进一步进行特征融合。
在一个实施例中，特征融合过程包括：
步骤S131，对ECG信号特征提取后得到的特征向量组成第一特征的样本空间X∈Rp×n。
例如，ECG信号特征反映波群形态特征和间期特征等。心电信号特征包括心电信号的时域特征、心电信号的变换域特征等。
步骤S132，对PPG信号特征提取后得到的特征向量矩阵组成第二特征的样本空间Y∈Rq×n。
步骤S133，采用最大最小值法，对特征向量进行归一化操作，使得两组特征向量的数量级接近。
优选地，通过对特征向量进行归一化处理，以消除奇异样本数据对后续预测结果的不良影响。
步骤S140，利用融合特征获得心脏病分类结果。
在该步骤中，利用融合特征训练分类器，进而利用经训练的分类器进行实时心脏病诊断分析。
在一个实施例中，选用支持向量机(SVM)作为分类器。这是因为SVM是定义在特征空间上的间隔最大的分类器，并更适用于小样本学习，并且SVM最终的决策结果取决于少数支持向量，从而可以“剔除”大量冗余样本，计算效率较高并具有较好的“鲁棒”性。由于大量心脏病患者的心电图样本和PPG信号样本可能不易获得，并且在实际应用中，患者的相关数据可能差异较大，因此选用SVM可获得更准确的诊断结果，如正常心跳、左心室动脉受阻心率、右心室动脉受阻心率、房颤、心室收缩和心房收缩等。
此外，为了进一步降维特征数据并保证后续诊断结果的准确性，在此步骤中，还可以利用主成分分析选择关键特征作为分类器的输入。
相应地，本发明还提供一种基于机器学习的非侵入性心脏病诊断装置，用于实现上述方法的一个方面或多个方面。例如，该装置包括：第一数据采集单元，其用于采集目标的心电图信号并检测对应的QRS波群；第二数据采集单元，其用于采集该目标的脉搏波信号；第一特征提取单元，其用于基于所述QRS波群提取心电图信号特征，构建为第一特征向量的样本空间；第二特征提取单元，其用于提取所述脉搏波信号的特征，将得到的特征向量矩阵构建为第二特征向量的样本空间；分类识别单元，其用于将所述第一特征向量的样本空间和所述第二特征向量的样本空间进行融合，利 用融合特征输入到预训练的分类器，获得心脏诊断结果。该装置所涉及的各单元可采用处理器、FPGA或专用硬件实现。
综上，本发明使用医院患者的数据，分别收集患者的脉搏波信号和心电图，首先利用LSTM提取PPG信号的特征。其次利用Pan-Tompkins算法，使用收集的心电图数据集对于心电信号QRS波群进行特征提取，在截裁样本后使用K-Means聚类进行数据筛选。再其次将提取的PPG信号特征和ECG信号特征进行融合，最后利用分类器对融合的特征进行分类。本发明提出了一种基于ECG和PPG相结合的新型心脏识别方法，首次将ECG和PPG相结合，为心脏识别提供了补充的解决方案，并提高了利用机器学习对心脏识别的准确率，获得的心脏诊断分类结果可用于临床指示。
本发明可以是系统、方法和/或计算机程序产品。计算机程序产品可以包括计算机可读存储介质，其上载有用于使处理器实现本发明的各个方面的计算机可读程序指令。
计算机可读存储介质可以是可以保持和存储由指令执行设备使用的指令的有形设备。计算机可读存储介质例如可以是但不限于电存储设备、磁存储设备、光存储设备、电磁存储设备、半导体存储设备或者上述的任意合适的组合。计算机可读存储介质的更具体的例子(非穷举的列表)包括：便携式计算机盘、硬盘、随机存取存储器(RAM)、只读存储器(ROM)、可擦式可编程只读存储器(EPROM或闪存)、静态随机存取存储器(SRAM)、便携式压缩盘只读存储器(CD-ROM)、数字多功能盘(DVD)、记忆棒、软盘、机械编码设备、例如其上存储有指令的打孔卡或凹槽内凸起结构、以及上述的任意合适的组合。这里所使用的计算机可读存储介质不被解释为瞬时信号本身，诸如无线电波或者其他自由传播的电磁波、通过波导或其他传输媒介传播的电磁波(例如，通过光纤电缆的光脉冲)、或者通过电线传输的电信号。
这里所描述的计算机可读程序指令可以从计算机可读存储介质下载到各个计算/处理设备，或者通过网络、例如因特网、局域网、广域网和/或无线网下载到外部计算机或外部存储设备。网络可以包括铜传输电缆、光纤传输、无线传输、路由器、防火墙、交换机、网关计算机和/或边缘服 务器。每个计算/处理设备中的网络适配卡或者网络接口从网络接收计算机可读程序指令，并转发该计算机可读程序指令，以供存储在各个计算/处理设备中的计算机可读存储介质中。
用于执行本发明操作的计算机程序指令可以是汇编指令、指令集架构(ISA)指令、机器指令、机器相关指令、微代码、固件指令、状态设置数据、或者以一种或多种编程语言的任意组合编写的源代码或目标代码，所述编程语言包括面向对象的编程语言—诸如Smalltalk、C++、Python等，以及常规的过程式编程语言—诸如“C”语言或类似的编程语言。计算机可读程序指令可以完全地在用户计算机上执行、部分地在用户计算机上执行、作为一个独立的软件包执行、部分在用户计算机上部分在远程计算机上执行、或者完全在远程计算机或服务器上执行。在涉及远程计算机的情形中，远程计算机可以通过任意种类的网络—包括局域网(LAN)或广域网(WAN)—连接到用户计算机，或者，可以连接到外部计算机(例如利用因特网服务提供商来通过因特网连接)。在一些实施例中，通过利用计算机可读程序指令的状态信息来个性化定制电子电路，例如可编程逻辑电路、现场可编程门阵列(FPGA)或可编程逻辑阵列(PLA)，该电子电路可以执行计算机可读程序指令，从而实现本发明的各个方面。
这里参照根据本发明实施例的方法、装置(系统)和计算机程序产品的流程图和/或框图描述了本发明的各个方面。应当理解，流程图和/或框图的每个方框以及流程图和/或框图中各方框的组合，都可以由计算机可读程序指令实现。
这些计算机可读程序指令可以提供给通用计算机、专用计算机或其它可编程数据处理装置的处理器，从而生产出一种机器，使得这些指令在通过计算机或其它可编程数据处理装置的处理器执行时，产生了实现流程图和/或框图中的一个或多个方框中规定的功能/动作的装置。也可以把这些计算机可读程序指令存储在计算机可读存储介质中，这些指令使得计算机、可编程数据处理装置和/或其他设备以特定方式工作，从而，存储有指令的计算机可读介质则包括一个制造品，其包括实现流程图和/或框图中的一个或多个方框中规定的功能/动作的各个方面的指令。
也可以把计算机可读程序指令加载到计算机、其它可编程数据处理装置、或其它设备上，使得在计算机、其它可编程数据处理装置或其它设备上执行一系列操作步骤，以产生计算机实现的过程，从而使得在计算机、其它可编程数据处理装置、或其它设备上执行的指令实现流程图和/或框图中的一个或多个方框中规定的功能/动作。
附图中的流程图和框图显示了根据本发明的多个实施例的系统、方法和计算机程序产品的可能实现的体系架构、功能和操作。在这点上，流程图或框图中的每个方框可以代表一个模块、程序段或指令的一部分，所述模块、程序段或指令的一部分包含一个或多个用于实现规定的逻辑功能的可执行指令。在有些作为替换的实现中，方框中所标注的功能也可以以不同于附图中所标注的顺序发生。例如，两个连续的方框实际上可以基本并行地执行，它们有时也可以按相反的顺序执行，这依所涉及的功能而定。也要注意的是，框图和/或流程图中的每个方框、以及框图和/或流程图中的方框的组合，可以用执行规定的功能或动作的专用的基于硬件的系统来实现，或者可以用专用硬件与计算机指令的组合来实现。对于本领域技术人员来说公知的是，通过硬件方式实现、通过软件方式实现以及通过软件和硬件结合的方式实现都是等价的。
以上已经描述了本发明的各实施例，上述说明是示例性的，并非穷尽性的，并且也不限于所披露的各实施例。在不偏离所说明的各实施例的范围和精神的情况下，对于本技术领域的普通技术人员来说许多修改和变更都是显而易见的。本文中所用术语的选择，旨在最好地解释各实施例的原理、实际应用或对市场中的技术改进，或者使本技术领域的其它普通技术人员能理解本文披露的各实施例。本发明的范围由所附权利要求来限定。

Claims (10)

  1. 一种基于机器学习的非侵入性心脏病诊断方法，包括以下步骤：
    采集目标的心电图信号并检测对应的QRS波群；
    采集该目标的脉搏波信号；
    基于所述QRS波群提取心电图信号特征，构建为第一特征向量的样本空间；
    提取所述脉搏波信号的特征，将得到的特征向量矩阵构建为第二特征向量的样本空间；
    将所述第一特征向量的样本空间和所述第二特征向量的样本空间进行融合，利用融合特征输入到预训练的分类器，获得心脏诊断结果。
  2. 根据权利要求1所述的方法，其特征在于，所述采集目标的心电图信号并检测对应的QRS波群包括：
    对于采集的心电图信号，利用Pan-Tompkins算法进行R波检测；
    基于检测到的R波对信号进行切割，获得RR间期特征，并在截裁样本使用聚类算法进行数据筛选。
  3. 根据权利要求1所述的方法，其特征在于，所述脉搏波信号的特征利用长短时记忆网络获得。
  4. 根据权利要求1所述的方法，其特征在于，所述分类器是支持向量机。
  5. 根据权利要求1所述的方法，其特征在于，在将融合特征输入分类器之前还包括：对所述第一特征向量的样本空间和所述第二特征向量的样本空间进行归一化处理。
  6. 根据权利要求1所述的方法，其特征在于，检测对应的QRS波群包括：
    分别采用低通滤波器和高通滤波器对采集的心电图信号进行滤波，其中低通滤波器的截止频率设置为11Hz，高通滤波器的截至频率设置为2Hz；
    利用微分方法对于QRS波群进行陡坡特性的放大并通过非线性平方函数增强心电图信号中的高频特性；
    利用移动窗积分确定波群中R波之外的其他特征；
    利用阈值动态调整和搜索确定心电图信号的峰值。
  7. 根据权利要求1所述的方法，其特征在于，所述脉搏波信号利用可穿戴设备基于光电容积脉搏波描记法获得。
  8. 根据权利要求2所述的方法，其特征在于，所述聚类算法是K-Means聚类。
  9. 一种基于机器学习的非侵入性心脏病诊断装置，包括：
    第一数据采集单元：用于采集目标的心电图信号并检测对应的QRS波群；
    第二数据采集单元：用于采集该目标的脉搏波信号；
    第一特征提取单元：用于基于所述QRS波群提取心电图信号特征，构建为第一特征向量的样本空间；
    第二特征提取单元：用于提取所述脉搏波信号的特征，将得到的特征向量矩阵构建为第二特征向量的样本空间；
    分类识别单元：用于将所述第一特征向量的样本空间和所述第二特征向量的样本空间进行融合，利用融合特征输入到预训练的分类器，获得心脏诊断结果。
  10. 一种计算机可读存储介质，其上存储有计算机程序，其中，该程序被处理器执行时实现根据权利要求1至8中任一项所述方法的步骤。
PCT/CN2021/137920 2021-11-11 2021-12-14 一种基于机器学习的非侵入性心脏病诊断方法和装置 WO2023082402A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202111331695.8A CN114081502B (zh) 2021-11-11 2021-11-11 一种基于机器学习的非侵入性心脏病诊断方法和装置
CN202111331695.8 2021-11-11

Publications (1)

Publication Number Publication Date
WO2023082402A1 true WO2023082402A1 (zh) 2023-05-19

Family

ID=80299785

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2021/137920 WO2023082402A1 (zh) 2021-11-11 2021-12-14 一种基于机器学习的非侵入性心脏病诊断方法和装置

Country Status (2)

Country Link
CN (1) CN114081502B (zh)
WO (1) WO2023082402A1 (zh)

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103284702A (zh) * 2013-06-17 2013-09-11 中国科学院苏州纳米技术与纳米仿生研究所 心电图和脉搏波关系分析及融合分析的方法和装置
CN109497992A (zh) * 2019-01-04 2019-03-22 济南汇医融工科技有限公司 基于机器学习方法的冠心病智能筛查装置
US20200093389A1 (en) * 2018-09-24 2020-03-26 Sotera Wireless, Inc. Method and system for monitoring a patient for atrial fibrillation and/or asystole
CN111358450A (zh) * 2020-03-17 2020-07-03 乐普（北京）医疗器械股份有限公司 一种血压分类方法和装置
CN111839505A (zh) * 2020-07-02 2020-10-30 复旦大学 一种基于心血管系统电-机械活动信息的房颤分级方法

Family Cites Families (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104706349B (zh) * 2015-04-13 2017-03-22 大连理工大学 一种基于脉搏波信号的心电信号构建方法
US10123741B2 (en) * 2016-11-30 2018-11-13 Huami Inc. Cardiac condition detection
EP3381356A1 (en) * 2017-03-28 2018-10-03 IMEC vzw System and method for cuffless blood pressure estimation
EP3449820B1 (en) * 2017-08-30 2023-05-24 Qompium Computer-implemented method and system for direct photoplethysmography (ppg)
CN111657925A (zh) * 2020-07-08 2020-09-15 中国科学院苏州生物医学工程技术研究所 基于机器学习的心电信号分类方法、系统、终端以及存储介质
CN112869717B (zh) * 2021-02-25 2023-02-24 佛山科学技术学院 一种基于bl-cnn的脉搏特征识别分类系统及方法
CN113349753A (zh) * 2021-07-19 2021-09-07 成都芯跳医疗科技有限责任公司 一种基于便携式动态心电监护仪的心律失常检测方法

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103284702A (zh) * 2013-06-17 2013-09-11 中国科学院苏州纳米技术与纳米仿生研究所 心电图和脉搏波关系分析及融合分析的方法和装置
US20200093389A1 (en) * 2018-09-24 2020-03-26 Sotera Wireless, Inc. Method and system for monitoring a patient for atrial fibrillation and/or asystole
CN109497992A (zh) * 2019-01-04 2019-03-22 济南汇医融工科技有限公司 基于机器学习方法的冠心病智能筛查装置
CN111358450A (zh) * 2020-03-17 2020-07-03 乐普（北京）医疗器械股份有限公司 一种血压分类方法和装置
CN111839505A (zh) * 2020-07-02 2020-10-30 复旦大学 一种基于心血管系统电-机械活动信息的房颤分级方法

Also Published As

Publication number Publication date
CN114081502B (zh) 2022-09-27
CN114081502A (zh) 2022-02-25

Similar Documents

Publication Publication Date Title
Vijayakumar et al. Fusion based feature extraction analysis of ECG signal interpretation–a systematic approach
Devi et al. Machine learning and IoT-based cardiac arrhythmia diagnosis using statistical and dynamic features of ECG
Bhagyalakshmi et al. GB-SVNN: Genetic BAT assisted support vector neural network for arrhythmia classification using ECG signals
Zhang et al. Localization of myocardial infarction with multi-lead bidirectional gated recurrent unit neural network
Sumathi et al. A wavelet transform based feature extraction and classification of cardiac disorder
Sedighian et al. Pediatric heart sound segmentation using Hidden Markov Model
Gupta Wavelet transform and vector machines as emerging tools for computational medicine
CN113095302A (zh) 用于心律失常分类的深度模型、利用该模型的方法及装置
Wu et al. Personalizing a generic ECG heartbeat classification for arrhythmia detection: a deep learning approach
Lu et al. An arrhythmia classification algorithm using C-LSTM in physiological parameters monitoring system under internet of health things environment
Li et al. A particle swarm optimization improved BP neural network intelligent model for electrocardiogram classification
Rawi et al. Classification and detection of ECG arrhythmia and myocardial infarction using deep learning: a review
He et al. Localization of myocardial infarction using a multi-branch weight sharing network based on 2-D vectorcardiogram
Liaqat et al. A review and comparison of the state-of-the-art techniques for atrial fibrillation detection and skin hydration
WO2023082402A1 (zh) 一种基于机器学习的非侵入性心脏病诊断方法和装置
Hong et al. Gated temporal convolutional neural network and expert features for diagnosing and explaining physiological time series: a case study on heart rates
Monroy et al. Hidden Markov model-based heartbeat detector using different transformations of ECG and ABP signals
Heng et al. Prediction of ventricular fibrillation using support vector machine
Zhang et al. Classification of pulmonary arterial pressure using photoplethysmography and bi-directional LSTM
Maghawry et al. An efficient approach for paroxysmal atrial fibrillation events prediction using extreme learning machine
Hasan et al. Cardiac arrhythmia detection in an ECG beat signal using 1D convolution neural network
Wadhwani et al. IOT based biomedical wireless sensor networks and machine learning algorithms for detection of diseased conditions
Liu et al. Adaptive atrial fibrillation detection focused on atrial activity analysis
Kaliappan et al. Automatic ECG analysis system with hybrid optimization algorithm based feature selection and classifier
Babu et al. Machine Learning Techniques Applied for Exploring Heart Disease and Classifying Stages through ECG Signal

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: 21963840

Country of ref document: EP

Kind code of ref document: A1