WO2019024861A1 - 心电信号的检测 - Google Patents

心电信号的检测 Download PDF

Info

Publication number
WO2019024861A1
WO2019024861A1 PCT/CN2018/097917 CN2018097917W WO2019024861A1 WO 2019024861 A1 WO2019024861 A1 WO 2019024861A1 CN 2018097917 W CN2018097917 W CN 2018097917W WO 2019024861 A1 WO2019024861 A1 WO 2019024861A1
Authority
WO
WIPO (PCT)
Prior art keywords
set number
neural network
convolutional neural
heart beat
single heart
Prior art date
Application number
PCT/CN2018/097917
Other languages
English (en)
French (fr)
Inventor
汪孔桥
赵威
李亚钊
李潇
Original Assignee
安徽华米信息科技有限公司
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 安徽华米信息科技有限公司 filed Critical 安徽华米信息科技有限公司
Priority to KR1020207005872A priority Critical patent/KR102451795B1/ko
Priority to EP18841762.0A priority patent/EP3644220A4/en
Priority to JP2020526667A priority patent/JP7065185B2/ja
Priority to CA3071699A priority patent/CA3071699C/en
Publication of WO2019024861A1 publication Critical patent/WO2019024861A1/zh
Priority to US16/750,568 priority patent/US11534097B2/en

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/217—Validation; Performance evaluation; Active pattern learning techniques
    • 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
    • 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
    • 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]
    • 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
    • A61B5/364—Detecting abnormal ECG interval, e.g. extrasystoles, ectopic heartbeats
    • 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
    • A61B5/7271—Specific aspects of physiological measurement analysis
    • A61B5/7275—Determining trends in physiological measurement data; Predicting development of a medical condition based on physiological measurements, e.g. determining a risk factor
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
    • G06F18/24133—Distances to prototypes
    • G06F18/24137—Distances to cluster centroïds
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
    • G16H50/20—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for computer-aided diagnosis, e.g. based on medical expert systems
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/12—Classification; Matching
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
    • Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
    • Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation

Definitions

  • the present application relates to the field of electronic technologies, and in particular, to a method, device, and electronic device for detecting an electrocardiogram signal.
  • the present application provides a new technical solution for pathological diagnosis of continuous ECG signals.
  • a method for detecting an electrocardiogram signal including:
  • a method for detecting an electrocardiogram signal including:
  • the ECG signal of the set time length is segmented to obtain a first set number of single heart beats
  • the first set number of single-heart beat data is input to the first convolutional neural network to determine, by the first convolutional neural network, an abnormal heart beat position in the first set number of single-heart beats.
  • an apparatus for detecting an electrocardiogram signal includes:
  • a first dividing module configured to divide the ECG signal of the set time length to obtain a first set number of single heart beats
  • a first determining module configured to determine feature data corresponding to each single heart beat of the first set number of single heart beats obtained by the first splitting module, to obtain feature data of a first set number
  • a second determining module configured to determine an ECG signal of the set time length based on the ECG signal of the set time length and the first set number of feature data determined by the first determining module Pathological category.
  • an apparatus for detecting an electrocardiogram signal comprising:
  • a fourth determining module configured to determine, by using the second convolutional neural network, a pathological category of the ECG signal of a set time length
  • a second segmentation module configured to: if the pathological category determined by the fourth determining module indicates that the ECG signal is abnormal, segment the ECG signal of the set time length to obtain a first set number Single heart beat;
  • a fifth determining module configured to input data of the first set number of single-heart beats obtained by the second splitting module to a first convolutional neural network, to determine the first by using a first convolutional neural network An abnormal heart beat position occurs in a single heart beat set.
  • a machine readable storage medium the storage medium storing machine executable instructions for performing the electrocardiogram of the first aspect or the second aspect described above Signal detection method.
  • an electronic device characterized in that the device comprises:
  • a storage medium for storing the processor-executable instructions
  • the processor is configured to perform the method for detecting an electrocardiogram signal according to the first aspect or the second aspect.
  • any single heart beat is not isolated in the continuous timing ECG signal, it is related to the single heart beat adjacent to it, so the first set number of single heart beats in the present application is single
  • the characteristic data corresponding to the heart beat can be used to well characterize the pathological characteristics of the representative electrocardiographic signal, so that the set time length of the electrocardiogram can be well detected by the ECG signal and the first set number of characteristic data.
  • the pathological category of the signal is the pathological category of the signal.
  • FIG. 1A is a schematic flow chart of a method for detecting an electrocardiogram signal according to an exemplary embodiment of the present invention.
  • Figure 1B is a schematic illustration of a continuous electrocardiographic signal in the embodiment of Figure 1A.
  • Figure 1C is a schematic illustration of a single heart beat in the embodiment of Figure 1A.
  • FIG. 2A is a schematic flow chart of a method for detecting an electrocardiogram signal according to another exemplary embodiment of the present invention.
  • FIG. 2B is a schematic structural diagram of detecting an electrocardiogram signal applied to the embodiment shown in FIG. 2A.
  • FIG. 2C is a schematic structural view of a first convolutional neural network in the embodiment shown in FIG. 2A.
  • FIG. 2C is a schematic structural view of a second convolutional neural network in the embodiment shown in FIG. 2A.
  • FIG. 3A is a schematic flow chart of a method for detecting an electrocardiogram signal according to still another exemplary embodiment of the present invention.
  • FIG. 3B is one of the schematic diagrams of the structure for detecting an electrocardiogram signal applied to the embodiment shown in FIG. 3A.
  • FIG. 3C is a second schematic structural diagram of detecting an electrocardiogram signal applied to the embodiment shown in FIG. 3A.
  • 4A is a schematic flow chart of a method for detecting an electrocardiogram signal according to still another exemplary embodiment of the present invention.
  • FIG. 4B is a schematic structural diagram of detecting an electrocardiogram signal applied to the embodiment shown in FIG. 4A.
  • FIG. 5 is a schematic structural diagram of an apparatus for detecting an electrocardiogram according to an exemplary embodiment of the present invention.
  • FIG. 6 is a schematic structural diagram of an apparatus for detecting an electrocardiogram according to another exemplary embodiment of the present invention.
  • FIG. 7 is a schematic structural diagram of an apparatus for detecting an electrocardiogram according to still another exemplary embodiment of the present invention.
  • FIG. 8 is a schematic structural diagram of an electronic device according to an exemplary embodiment of the present invention.
  • first, second, third, etc. may be used to describe various information in this application, such information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other.
  • first information may also be referred to as the second information without departing from the scope of the present application.
  • second information may also be referred to as the first information.
  • word "if” as used herein may be interpreted as "when” or "when” or "in response to determination.”
  • FIG. 1A is a schematic flowchart of a method for detecting an electrocardiogram signal according to an exemplary embodiment of the present invention.
  • FIG. 1B is a schematic diagram of a continuous electrocardiographic signal in the embodiment shown in FIG. 1A
  • FIG. 1C is a schematic diagram of the embodiment shown in FIG. 1A.
  • step 101 the ECG signal of the set time length is divided to obtain a first set number of single heart beats.
  • the ECG signal of the set time length may be segmented by the method for recognizing the ECG signal to obtain a first set number of single-heart beats.
  • the start time point, the end time point, and the duration of each single heart beat can be known, and the duration t of the single heart beat can be unified.
  • the single heart beat can be expressed as:
  • the sampling may be performed according to a fixed sampling rate
  • P represents the signal strength of the single heart beat at each sampling point
  • P 1 represents a single heart beat at the first sample.
  • P L represents the signal strength of the single heart beat at the Lth sample point
  • p L represents the signal strength of the single heart beat at the L sample points
  • t is the duration of the single heart beat
  • the duration is described by taking the first set number as N.
  • the continuous N ECG single heart beats can be expressed as:
  • p ij represents the signal strength of the i-th single-heart beat at the j-th sample point
  • Step 102 Determine feature data corresponding to each single heart beat of the first set number of single heart beats to obtain feature data of the first set number.
  • the feature data corresponding to each single heart beat of the first set number of single heart beats may be determined based on the deep learning network, and the feature data may be a single feature or a combination of multiple features.
  • the single-heart beat signal can be input into the deep learning network, and the feature data is obtained by convoluting the single-heart beat signal by setting the convolution layer of the deep learning network.
  • the feature data is, for example, T 1 , T 2 , ..., T N .
  • Step 103 Determine a pathological category of the ECG signal of a set time length based on the ECG signal of the set time length and the first set number of feature data.
  • the set time length ECG signal may be input to an input layer of a convolutional neural network, and the first set number of feature data may be input to the convolution layer of the convolutional neural network.
  • the processing of the convolutional neural network determines the pathological category of the ECG signal at the output layer of the convolutional neural network.
  • the convolutional neural network can be trained by a large number of ECG signals having various pathological features. By training the convolutional neural network, the convolutional neural network can accurately identify the ECG. The pathological category of the signal.
  • the pathological category may include: atrial premature beats, ventricular premature beats, atrial fibrillation, atrial flutter, supraventricular tachycardia, etc., it should be noted that the above pathological categories are merely exemplary and cannot form a pair. Application restrictions.
  • any single heart beat in the continuous ECG signal is not isolated, and is related to the single heart beat adjacent to the front and rear
  • the feature data corresponding to each single heart beat of the first set number of single heart beats in the present application can be used.
  • the pathological characteristics of the representative ECG signal are well characterized, so that the pathological category of the ECG signal of the set time length can be well detected by the ECG signal and the first set number of characteristic data.
  • FIG. 2A is a schematic flow chart of a method for detecting an electrocardiogram signal according to another exemplary embodiment of the present invention
  • FIG. 2B is a schematic structural diagram of detecting an electrocardiogram signal applied to the embodiment shown in FIG. 2A
  • FIG. 2C is a schematic diagram of the implementation shown in FIG. 2A
  • FIG. 2D is a schematic structural diagram of a second convolutional neural network in the embodiment shown in FIG. 2A; as shown in FIG. 2A, the following steps are included:
  • step 201 the ECG signal of the set time length is divided to obtain a first set number of single heart beats.
  • step 201 For the description of step 201, reference may be made to the description of the embodiment shown in FIG. 1A above, and details are not described herein again.
  • Step 202 sequentially input the first set number of single-heart beat data into the first convolutional neural network to extract feature data corresponding to each single heart beat of the first set number of single heart beats.
  • Step 203 Determine time series data corresponding to each single heart beat of the ECG signals of the set time length, and obtain time series data of the first set number.
  • Step 204 Input the first set number of time series data and the first set number of feature data into an input layer of the second convolutional neural network to determine a pathological category of the ECG signal.
  • Step 205 Determine, by using the first convolutional neural network, each single heart beat of the first set number of single heart beats to obtain a judgment result.
  • Step 206 Determine, according to the determination result, an abnormal heart beat position in the first set number of single heart beats.
  • steps 205 and 206 are not necessarily performed after the step 204, and after step 202, the steps 205 and 206 are performed, so that the first set may be identified by the first convolutional neural network.
  • An abnormal heart beat position appears in the single heart beat of the number.
  • the continuous 25 single-heart beats and the start time point, the end time point, and the duration of the 25 single-heart beats are obtained through the above step 201, and the length of each single-heart beat data is performed.
  • Normalization if the normalization length is 196, 25 single-heart beat data of length 196 can be sequentially input to the first convolutional neural network, in the first convolutional layer of the first convolutional neural network, The feature data corresponding to each single heart beat is output.
  • each single heart beat corresponds to 5*2 feature data
  • 25 single heart beats can correspond to 5*2*25 feature data.
  • the feature data corresponding to each single heart beat can be cached.
  • the time point corresponding to the R wave of the single heart beat may be determined for each single heart beat of the set of time ECG signals, and the single heart beat is determined. a time point corresponding to each of the R waves corresponding to the second set number of single-heart beats before and after the R wave, a time point corresponding to the R wave based on the single heart beat, and a second set number adjacent to the single heart beat The time points corresponding to the R waves of the heart beat are determined, and the time series data corresponding to the single heart beat is determined.
  • the R wave distances of the current single heart beat R wave and the front and rear two single heart beats are x 1 , x 2 , x 3 , x 4 , respectively.
  • the rhythm information X of each single heart beat is represented as a 5-dimensional vector, that is,
  • N single heart beats can be expressed as:
  • N is the first set number, that is, the number of single heart beats.
  • the amount of data input to the second convolutional neural network may be expressed as m1*n1+m2*n2.
  • the single heart beat input to the second convolutional neural network corresponds to 5*2*25 feature data
  • the single heart beat input to the second convolutional neural network corresponds to 5*1*25 time series data, and then input.
  • the amount of data to the second convolutional neural network can be expressed as 5*3*25.
  • the second convolutional neural network when the second convolutional neural network is trained by a large number of ECG signals having various pathological features, the second convolutional neural network can accurately identify the ECG signal through the data of the input layer. Pathological category.
  • the first convolutional neural network can be trained by a large number of normal and abnormal single-heart beat ECG signals. After training, the first convolutional neural network can accurately recognize the single heart beat. Whether an exception has occurred. For example, when 25 single-heart beats are sequentially input to the first convolutional neural network, the first convolutional neural network can judge the normality and abnormality of the single-heart beat, for example, 1 means that the single heart beat is normal, and 0 means single heart beat. Abnormally, 25 single-heart beats can correspond to 25 combinations of 0 and 1 sequences. By combining the sequences, the heart beat positions in the 25 single-heart beats can be identified.
  • the first convolutional neural network includes four convolutional layers, each of which includes convolution, activation, and pooling.
  • the processing order of each processing is the same, that is, the data of the single heart beat is first convolved and activated after inputting the first convolutional neural network, and then Pooling, thereby obtaining characteristic data of the output of the convolution layer, wherein the first layer of the convolution layer to the third layer of the convolution layer, each layer comprises a continuous two-level convolution, activation, pooling, fourth layer
  • the convolutional layer has only one level of convolution, activation, and pooling.
  • the activation operation generally uses functions such as sigmoid, tanH, and reLu.
  • the convolution operation is used to extract the characteristic data of the ECG signal; the activation operation is used to improve the nonlinearity of the feature data, that is, the activity degree; the pooling operation is to reduce the dimension of the feature data.
  • the second convolutional neural network includes three convolutional layers. For each convolutional layer, except for the input and output data and the convolution kernel size, the processing order of each processing is the same, that is, input data ( After inputting the second convolutional neural network, the convolution and activation are performed to obtain characteristic data of the convolutional layer output, wherein the first convolutional layer to the third convolutional layer Each layer contains a first-level convolution and activation, and does not contain pooling.
  • the abnormal concentric beat position can be accurately given by the first convolutional neural network;
  • the output feature data can be regarded as an approximation of the original single-heart beat, so the feature data output by the first convolutional neural network is added to the original ECG signal, greatly enhancing the possibility of identifying the ECG signal.
  • FIG. 3A is a schematic flowchart of a method for detecting an electrocardiogram signal according to still another exemplary embodiment of the present invention
  • FIG. 3B is a schematic structural diagram of detecting an ECG signal applied to the embodiment shown in FIG. 3A
  • FIG. 3C is a schematic diagram of FIG.
  • the schematic diagram of the architecture for detecting the ECG signal is applied to the embodiment; as shown in FIG. 3A, the following steps are included:
  • Step 301 dividing the ECG signal of the set time length to obtain a first set number of single heart beats.
  • step 301 For the description of step 301, reference may be made to the description of the embodiment shown in FIG. 1A above, and details are not described herein again.
  • Step 302 sequentially input the first set number of single-heart beat data into the first convolutional neural network to extract feature data corresponding to each single heart beat of the first set number of single heart beats.
  • step 302 reference may be made to the description of the embodiment shown in FIG. 2A above, and details are not described herein again.
  • Step 303 input the first set number of feature data into the set convolution layer of the second convolutional neural network, and input the ECG signal of the set time length into the input layer of the second convolutional neural network to pass the The second convolutional neural network determines the pathological category of the ECG signal.
  • step 303 For the description of step 303, reference may be made to the description of the embodiment shown in FIG. 2A above, and details are not described herein again.
  • Step 303 is exemplarily described below in conjunction with FIGS. 3B and 3C.
  • the first set number of single-heart beat data obtained by the segmentation may be sequentially input to the first convolutional neural network, and the undivided original ECG signal is input to the second convolutional neural network.
  • the original continuous ECG signal is divided into 20 single-heart beats, and 20 single-heart beat data are sequentially input to the first convolutional neural network.
  • the characteristic data output from the set convolution layer of the first convolutional neural network may be buffered after the data of the 20 single-heart beats enters the first convolutional neural network and is processed accordingly.
  • the original continuous ECG signal is input to the input layer of the second convolutional neural network.
  • the feature data obtained by setting the convolutional layer of the first convolutional neural network may be injected into the feature data obtained by setting the convolution layer of the second convolutional neural network.
  • the feature data obtained by the second convolutional layer of the first convolutional neural network is injected into the feature data obtained by the second convolution layer of the second convolutional neural network, and the two groups are The feature data is input together to the third layer convolutional layer of the second convolutional neural network.
  • the amount of feature data obtained from the second convolutional layer of the second convolutional neural network is represented as 531*32
  • the amount of feature data that can be processed by the third layer convolutional layer is 531*64, for the third layer convolutional layer, 111 is missing. *32 feature data.
  • the feature data of the actual input of the third layer convolution layer can be ensured by the zero-padding method to be consistent with the feature data to be processed by the third layer convolution layer.
  • FIG. 3C is merely an exemplary illustration, and feature data of different dimensions may be output from a set convolution layer of the first convolutional neural network, and injected into a set convolution layer of the second convolutional neural network, thereby The degree of dependence of the second convolutional neural network on the feature data of the first convolutional neural network can be flexibly adjusted.
  • the set convolution layer here may be a convolutional layer or a plurality of convolutional layers.
  • the feature data output from the second convolutional layer of the first convolutional neural network may be injected into the second convolutional layer of the second convolutional neural network, and at the same time, the first convolutional neural network may be The feature data of the three-volume layer output is injected into the third convolutional layer of the second convolutional neural network.
  • the recognition result of the second convolutional neural network can be relied on
  • the characteristic data of a convolutional neural network can enhance the recognition performance of the ECG signal.
  • FIG. 4A is a schematic flowchart of a method for detecting an electrocardiogram signal according to still another exemplary embodiment of the present invention
  • FIG. 4B is a schematic structural diagram of detecting an electrocardiogram signal applied to the embodiment shown in FIG. 4A; as shown in FIG. 4A, including the following step:
  • Step 401 Input the ECG signal of the set time length into the second convolutional neural network and determine the pathological category of the ECG signal through the second convolutional neural network.
  • Step 402 If the pathological category indicates that the ECG signal is abnormal, the ECG signal of the set time length is divided to obtain a first set number of single heart beats.
  • Step 403 Input the obtained first set number of single-heart beat data into the first convolutional neural network, and determine, by the first convolutional neural network, the abnormally placed heart beat position in the first set number of single-heart beats. .
  • the data of the first set number of single-heart beats may be input to the input layer of the first convolutional neural network; and the first set number of the single set by the first convolutional neural network Each single heart beat in the heart beat is judged to obtain a judgment result; and according to the judgment result, an abnormal heart beat position in the first set number of single heart beats is determined.
  • data of a continuous electrocardiographic signal of a set length of time (for example, 30 seconds) is directly input to the second convolutional neural network, and the pathological category is obtained by identifying the data of the continuous electrocardiographic signal. If the pathological category indicates that the electrocardiographic signal is abnormal, the ECG signal can be segmented to obtain, for example, 25 single-heart beat data. The 25 single-hearted data are sequentially input to the first convolutional neural network to be identified, and the decision result is obtained. The result of the decision may be a sequence consisting of 0 and 1.
  • 0 means an abnormality
  • 1 means normal
  • 25 single-heart beats may correspond to a combination of 0 and 1 having a length of 25 bits, by identifying the position where 0 is located, that is, An abnormal heart beat position in these 25 single heart beats can be determined.
  • the single-heart beat of the ECG signal is identified one by one by the first convolutional neural network, thereby identifying an abnormality. Heart beat position.
  • the first convolutional neural network in the present application can be regarded as a single-heart beat recognition network
  • the second convolutional neural network can be regarded as an electrocardiogram signal.
  • the detection network, the architecture design of the first convolutional neural network and the second convolutional neural network in the present application may have the following beneficial technical effects:
  • the first convolutional neural network can be trained separately to reduce the difficulty of training the entire network architecture
  • the training of the second convolutional neural network can be enhanced based on the feature data obtained by the first convolutional neural network, thereby solving the problem of insufficient data of the ECG signal of the continuous heartbeat of the long sequence;
  • FIG. 5 is a schematic structural diagram of an apparatus for detecting an electrocardiogram signal according to an exemplary embodiment of the present invention. As shown in FIG. 5, the apparatus for detecting an electrocardiographic signal may include: a first segmentation module 51, a first determination module 52, and a first Two determining module 53. among them:
  • the first dividing module 51 is configured to divide the ECG signal of the set time length to obtain a first set number of single heart beats
  • the first determining module 52 is configured to determine feature data corresponding to each single heart beat of the first set number of single heart beats obtained by the first splitting module 51, to obtain feature data of the first set number;
  • the second determining module 53 is configured to determine a pathological category of the ECG signal of the set time length based on the ECG signal of the set time length and the first set number of feature data determined by the first determining module 52.
  • FIG. 6 is a schematic structural diagram of an apparatus for detecting an electrocardiogram signal according to another exemplary embodiment of the present invention.
  • the first determining module 52 may include:
  • a first input unit 521 configured to sequentially input data of the first set number of single-heart beats to the first convolutional neural network
  • the extracting unit 522 is configured to extract feature data corresponding to each single heart beat of the first set number of single heart beats by using the first convolutional neural network.
  • the apparatus for detecting an electrocardiogram signal further includes:
  • the determining module 54 is configured to determine, by using the first convolutional neural network, each single heart beat of the first set number of single heart beats obtained by the first splitting module 51, to obtain a determination result;
  • the third determining module 55 is configured to determine, according to the determination result obtained by the determining module 54, a heart beat position in which an abnormality occurs in the first set number of single heart beats.
  • the second determining module 53 can include:
  • the first determining unit 531 is configured to determine time series data corresponding to each single heart beat of the first set number of single heart beats, to obtain time data of the first set number;
  • the second input unit 532 is configured to input the first set number of time series data obtained by the first determining unit 531 and the first set number of feature data into the input layer of the second convolutional neural network;
  • the second determining unit 533 is configured to determine a pathological category of the ECG signal by using the second convolutional neural network.
  • the first determining unit 531 is specifically configured to:
  • the time series data corresponding to the single heart beat is determined based on the time point corresponding to the R wave of the single heart beat and the time point corresponding to the R wave of each of the second set number of single heart beats before and after the single heart beat.
  • the second determining module 53 comprises:
  • a third input unit 534 configured to input the first set number of feature data into a set convolution layer of the second convolutional neural network
  • a fourth input unit 535 configured to input an ECG signal of a set time length into an input layer of the second convolutional neural network
  • the third determining unit 536 is configured to identify the first set number of feature data input by the third input unit 534 and the ECG signal input by the fourth input unit 535 by using the second convolutional neural network to determine the ECG signal. Pathological category.
  • FIG. 7 is a schematic structural diagram of an apparatus for detecting an electrocardiogram signal according to still another exemplary embodiment of the present invention.
  • the apparatus for detecting an electrocardiogram signal may include: a fourth determining module 71, a second dividing module 72, and a Five determining module 73; wherein:
  • a fourth determining module 71 configured to determine, by using the second convolutional neural network, a pathological category of the ECG signal of a set time length
  • the second segmentation module 72 is configured to: if the pathological category determined by the fourth determining module 71 indicates that the ECG signal is abnormal, segment the ECG signal of the set time length to obtain a first set number of single heart beats;
  • the fifth determining module 73 is configured to input the first set number of single-heart beat data obtained by the second splitting module 72 to the first convolutional neural network, to determine the first set by the first convolutional neural network. An abnormal heart beat position appears in the single heart beat of the number.
  • the fifth determining module 73 can include:
  • a fifth input unit 731 configured to input data of the first set number of single heart beats into an input layer of the first convolutional neural network
  • the determining unit 732 is configured to determine, by the fifth input unit 731, the first convolutional neural network to determine each single heart beat of the first set number of single heart beats, to obtain a determination result;
  • the fourth determining unit 733 is configured to determine, according to the determination result obtained by the determining unit 732, a heart beat position in which an abnormality occurs in the first set number of single heart beats.
  • Embodiments of the memory detecting device of the present application can be applied to an electronic device.
  • the device embodiment may be implemented by software, or may be implemented by hardware or a combination of hardware and software.
  • the processor of the electronic device in which it is located reads a corresponding machine executable instruction in the non-volatile storage medium into the memory. From a hardware level, as shown in FIG.
  • the electronic device in which the device is located in the embodiment may also include other hardware according to the actual function of the electronic device, and details are not described herein.

Abstract

一种心电信号的检测方法、装置及电子设备。根据该方法的示例，对设定时间长度的心电信号进行分割，得到第一设定个数的单心拍（101）；确定所述第一设定个数的单心拍中每一个单心拍对应的特征数据，得到第一设定个数的特征数据（102）；基于所述设定时间长度的心电信号和所述第一设定个数的特征数据，确定所述设定时间长度的心电信号的病理类别（103）。

Description

心电信号的检测 技术领域
本申请涉及电子技术领域，尤其涉及一种心电信号的检测方法、装置及电子设备。
背景技术
近年来，随着深度学习方法的兴起，研究人员越来越多地开始采用训练神经网络的模式进行心电(ECG)信号分类和识别。但多是针对心电信号的单心拍进行识别并分类，而尚未对连续的ECG多心拍进行病理识别。
发明内容
有鉴于此，本申请提供一种新的技术方案，可以对连续的心电信号进行病理诊断。
为实现上述目的，本申请提供技术方案如下：
根据本申请的第一方面，提出了一种心电信号的检测方法，包括：
对设定时间长度的心电信号进行分割，得到第一设定个数的单心拍；
确定所述第一设定个数的单心拍中每一个单心拍对应的特征数据，得到第一设定个数的特征数据；
基于所述设定时间长度的心电信号和所述第一设定个数的特征数据，确定所述设定时间长度的心电信号的病理类别。
根据本申请的第二方面，提出了一种心电信号的检测方法，包括：
通过第二卷积神经网络确定设定时间长度的心电信号的病理类别；
若所述病理类别表示所述心电信号出现异常，对所述设定时间长度的心电信号进行分割，得到第一设定个数的单心拍；
将所述第一设定个数的单心拍的数据输入到第一卷积神经网络，以通过第一卷积神经网络确定所述第一设定个数的单心拍中出现异常的心拍位置。
根据本申请的第三方面，提供了一种心电信号的检测装置，包括：
第一分割模块，用于对设定时间长度的心电信号进行分割，得到第一设定个数的单心拍；
第一确定模块，用于确定所述第一分割模块得到的所述第一设定个数的单心拍中每一个单心拍对应的特征数据，得到第一设定个数的特征数据；
第二确定模块，用于基于所述设定时间长度的心电信号和所述第一确定模块确定的所述第一设定个数的特征数据，确定所述设定时间长度的心电信号的病理类别。
根据本申请的第四方面，提供了一种心电信号的检测装置，包括：
第四确定模块，用于通过第二卷积神经网络确定设定时间长度的心电信号的病理类别；
第二分割模块，用于若所述第四确定模块确定的所述病理类别表示所述心电信号出现异常，对所述设定时间长度的心电信号进行分割，得到第一设定个数的单心拍；
第五确定模块，用于将所述第二分割模块得到的所述第一设定个数的单心拍的数据输入到第一卷积神经网络，以通过第一卷积神经网络确定所述第一设定个数的单心拍中出现异常的心拍位置。
根据本申请的第五方面，提出了一种机器可读存储介质，所述存储介质存储有机器可执行指令，所述机器可执行指令用于执行上述第一方面或者第二方面提出的心电信号的检测方法。
根据本申请的第六五方面，提出了一种电子设备，其特征在于，所述设备包括：
处理器；用于存储所述处理器可执行指令的存储介质；
其中，所述处理器，用于执行上述第一方面或者第二方面提出的心电信号的检测方法。
由以上技术方案可见，由于任何单个心拍在连续时序的ECG信号中都不是孤立的，都与其前后相邻的单心拍相关，因此本申请中的第一设定个数的单心拍中每一个单心拍对应的特征数据可以用来很好地表征所代表的心电信号的病理特性，因此通过心电信号和第一设定个数的特征数据能够很好地检测出设定时间长度的心电信号的病理类别。
附图说明
图1A是本发明一示例性实施例的心电信号的检测方法的流程示意图。
图1B是图1A所示实施例中的连续心电信号的示意图。
图1C是图1A所示实施例中的单心拍的示意图。
图2A是本发明另一示例性实施例的心电信号的检测方法的流程示意图。
图2B是图2A所示实施例所适用的检测心电信号的架构示意图。
图2C是图2A所示实施例中的第一卷积神经网络的结构示意图。
图2C是图2A所示实施例中的第二卷积神经网络的结构示意图。
图3A是本发明又一示例性实施例的心电信号的检测方法的流程示意图。
图3B是图3A所示实施例所适用的检测心电信号的架构示意图之一。
图3C是图3A所示实施例所适用的检测心电信号的架构示意图之二。
图4A是本发明再一示例性实施例的心电信号的检测方法的流程示意图。
图4B是图4A所示实施例所适用的检测心电信号的架构示意图。
图5是本发明一示例性实施例的心电信号的检测装置的结构示意图。
图6是本发明另一示例性实施例的心电信号的检测装置的结构示意图。
图7是本发明又一示例性实施例的心电信号的检测装置的结构示意图。
图8是本发明一示例性实施例的电子设备的结构示意图。
具体实施方式
这里将详细地对示例性实施例进行说明，其示例表示在附图中。下面的描述涉及附图时，除非另有表示，不同附图中的相同数字表示相同或相似的要素。以下示例性实施例中所描述的实施方式并不代表与本申请相一致的所有实施方式。相反，它们仅是与如所附权利要求书中所详述的、本申请的一些方面相一致的装置和方法的例子。
在本申请使用的术语是仅仅出于描述特定实施例的目的，而非旨在限制本申请。在本申请和所附权利要求书中所使用的单数形式的“一种”、“所述”和“该”也旨在包括多数形式，除非上下文清楚地表示其他含义。还应当理解，本文中使用的术语“和/ 或”是指并包含一个或多个相关联的列出项目的任何或所有可能组合。
应当理解，尽管在本申请可能采用术语第一、第二、第三等来描述各种信息，但这些信息不应限于这些术语。这些术语仅用来将同一类型的信息彼此区分开。例如，在不脱离本申请范围的情况下，第一信息也可以被称为第二信息，类似地，第二信息也可以被称为第一信息。取决于语境，如在此所使用的词语“如果”可以被解释成为“在……时”或“当……时”或“响应于确定”。
为对本申请进行进一步说明，提供下列实施例：
图1A是本发明一示例性实施例的心电信号的检测方法的流程示意图，图1B是图1A所示实施例中的连续心电信号的示意图，图1C是图1A所示实施例中的ECG单心拍的示意图；本申请可应用在可穿戴设备以及手持设备等电子设备中，以监测用户的心脏健康状况，如图1A所示，包括如下步骤：
步骤101，对设定时间长度的心电信号进行分割，得到第一设定个数的单心拍。
在一实施例中，可以通过关于ECG信号的识别方法对设定时间长度的ECG信号进行分割，得到第一设定个数的单心拍。如图1B和图1C所示，通过对设定时间长度的连续ECG信号进行分割，可获知每个单心拍的开始时间点、结束时间点以及持续时长，且单心拍的持续时长t可以归一化为预设长度，例如，预设长度为L，则单心拍可以表示为：
e＝(P,X)＝((p 1,p 2,…,p L),t)
其中，在对单心拍的持续时长进行归一化的过程中，可以按照固定的采样率进行采样，P表示单心拍在各个采样点的信号强度，例如，P 1表示单个心拍在第1个采样点的信号强度，P L表示单个心拍在第L个采样点的信号强度，p 1,p 2,…,p L表示单心拍在L个采样点处的信号强度，t为该单心拍的持续时长，以第一设定个数为N进行示例性说明，连续N个ECG单心拍可表示为：
E＝{e 1,e 2,…,e N}＝{P 1,P 2,…,P N,t 1,t 2,…t N}
＝{(p 11,p 12,…,p 1L),(p 21,p 22,…,p 2L),…,(p N1,p N2,…,p NL),t 1,t 2,…,t N}。
其中，p ij表示第i个单心拍在第j处采样点的信号强度，i＝1,2,…,N，j＝1,2,…,L，t 1,t 2,…,t N表示N个单心拍各自的持续时长。
步骤102，确定第一设定个数的单心拍中每一个单心拍对应的特征数据，得到第一 设定个数的特征数据。
在一实施例中，可以基于深度学习网络确定第一设定个数的单心拍中每一个单心拍对应的特征数据，该特征数据可以为单个特征也可以为多个特征的组合。可以将单心拍信号输入到深度学习网络中，通过设定该深度学习网络的卷积层对单心拍信号进行卷积处理后得到特征数据。
与上述步骤101相对应，特征数据例如为：T 1,T 2,…,T N。
即步骤102可以实现：P i→T i，其中，i＝1,2,…,N。
步骤103，基于设定时间长度的心电信号和第一设定个数的特征数据，确定设定时间长度的心电信号的病理类别。
在一实施例中，可以将设定时间长度的心电信号输入到一个卷积神经网络的输入层，将第一设定个数的特征数据输入到该卷积神经网络的卷积层，通过卷积神经网络的处理，在该卷积神经网络的输出层确定心电信号的病理类别。在一实施例中，可以通过海量的具有各种不同病理特征的心电信号对该卷积神经网络进行训练，通过训练该卷积神经网络，可以使该卷积神经网络能够准确识别出心电信号的病理类别。
在一实施例中，病理类别可以包括：房性早搏、室性早搏、心房颤动、心房扑动、室上性心动过速等，需要说明的是，上述病理类别只是示例性说明并不能形成对本申请的限制。
由于连续ECG信号中的任何单个心拍都不是孤立的，都与其前后相邻的单心拍相关，因此本申请中的第一设定个数的单心拍中每一个单心拍对应的特征数据可以用来很好地表征所代表的心电信号的病理特性，因此通过心电信号和第一设定个数的特征数据能够很好地检测出设定时间长度的心电信号的病理类别。
图2A是本发明另一示例性实施例的心电信号的检测方法的流程示意图，图2B是图2A所示实施例所适用的检测心电信号的架构示意图，图2C是图2A所示实施例中的第一卷积神经网络的结构示意图，图2D是图2A所示实施例中的第二卷积神经网络的结构示意图；如图2A所示，包括如下步骤：
步骤201，对设定时间长度的心电信号进行分割，得到第一设定个数的单心拍。
步骤201的描述可以参见上述图1A所示实施例的描述，在此不再详述。
步骤202，将第一设定个数的单心拍的数据顺次输入到第一卷积神经网络，以提取 第一设定个数的单心拍中每一个单心拍对应的特征数据。
步骤203，确定设定时间长度的心电信号中每一个单心拍对应的时序数据，得到第一设定个数的时序数据。
步骤204，将第一设定个数的时序数据和第一设定个数的特征数据输入第二卷积神经网络的输入层，以确定该心电信号的病理类别。
步骤205，通过第一卷积神经网络对第一设定个数的单心拍中的每一个单心拍进行判决，得到判决结果。
步骤206，根据判决结果确定第一设定个数的单心拍中出现异常的心拍位置。
需要说明的是，上述步骤205和步骤206并不必然在步骤204之后执行，也可以在步骤202之后，再执行步骤205和步骤206，从而可以通过第一卷积神经网络识别第一设定个数的单心拍中出现异常的心拍位置。
下面结合图2B-图2D对本实施例进行示例性说明。
在上述步骤202中，如图2B所示，通过上述步骤201得到连续25个单心拍，以及该25个单心拍的开始时间点、结束时间点以及持续时长，对每个单心拍的数据进行长度归一化，如归一化长度为196，则可以将25个长度为196的单心拍数据顺次输入到第一卷积神经网络，在第一卷积神经网络的预设卷积层，可以输出每一个单心拍对应的特征数据，例如，每个单心拍对应5*2个特征数据，则25个单心拍可对应5*2*25个特征数据。如图2B所示，可以在通过第一卷积神经网络得到每一个单心拍的特征数据后，缓存每一个单心拍对应的特征数据。
在上述步骤203和步骤204中，在一实施例中，可以针对设定时间长度的心电信号中的每一个单心拍，确定该单心拍的R波对应的时间点，确定与该单心拍的R波前后相邻第二设定个数的单心拍各自的R波对应的时间点，基于该单心拍的R波对应的时间点以及与该单心拍前后相邻第二设定个数的单心拍的R波各自对应的时间点，确定该单心拍对应的时序数据。例如，第二设定个数为2，则当前单心拍的R波与前后两个单心拍(即共四个单心拍)的R波距离分别为x 1,x 2,x 3,x 4，将各单心拍的节律信息X表示为一个5维向量，即
Figure PCTCN2018097917-appb-000001
则N个单心拍可以表示为：
Figure PCTCN2018097917-appb-000002
其中，N为第一设定个数，即单心拍的个数。
例如，单心拍对应m1*n1个特征数据，以及m2*n2个时序数据，则输入到第二卷积神经网络的数据量可以表示为m1*n1+m2*n2。如图2B所示，输入到第二卷积神经网络的单心拍对应5*2*25个特征数据，输入到第二卷积神经网络的单心拍对应5*1*25个时序数据，则输入到第二卷积神经网络的数据量可以表示为5*3*25。
在上述步骤204中，当第二卷积神经网络通过海量的具有各种病理特征的心电信号进行训练后，第二卷积神经网络即可通过输入层的数据，准确识别出该心电信号的病理类别。
在上述步骤205和步骤206中，可以对第一卷积神经网络通过海量的正常与异常的单心拍的心电信号进行训练，通过训练后，第一卷积神经网络即可准确识别出单心拍是否出现异常。例如，当将25个单心拍顺次输入到第一卷积神经网络后，第一卷积神经网络可以对单心拍的正常与异常做出判决，例如，1表示单心拍正常，0表示单心拍异常，则25个单心拍可对应25个0和1的序列组合，通过该序列组合，即可识别出这25个单心拍中出现异常的心拍位置。
如图2C所示，第一卷积神经网络包括四个卷积层，每层卷积均包括卷积、激活和池化(pooling)等处理。对于每个卷积层，除输入输出数据和卷积核尺寸不同外，各处理的计算顺序均相同，即单心拍的数据在输入第一卷积神经网络后，先经过卷积和激活，然后池化，从而得到该卷积层输出的特征数据，其中，第一层卷积层到第三层卷积层中，每层都包含连续的两级卷积、激活、池化，第四层卷积层只有一级卷积、激活、池化。其中，激活运算一般利用sigmoid、tanH、reLu等函数。在第一卷积神经网络中，卷积运算用以提取心电信号的特征数据；激活运算用以提升特征数据的非线性度，即活跃度；池化运算是对特征数据降维。
如图2D所示，第二卷积神经网络包括三个卷积层，对于每个卷积层，除输入输出数据和卷积核尺寸不同外，各处理的计算顺序均相同，即输入数据(包括心电信号的时序数据)在输入第二卷积神经网络后，经过卷积和激活，从而得到该卷积层输出的特征 数据，其中，第一层卷积层到第三层卷积层中，每层都包含一级卷积和激活，并未包含池化。
本实施例中，在第二卷积神经网络对连续的心电信号做出病理诊断的同时，还能够通过第一卷积神经网络准确地给出异常的心拍位置；由于第一卷积神经网络输出的特征数据可以看成是原始的单心拍的近似，因此将第一卷积神经网络输出的特征数据累加到原始的心电信号中，大大增强识别心电信号的可能性。
图3A是本发明又一示例性实施例的心电信号的检测方法的流程示意图，图3B是图3A所示实施例所适用的检测心电信号的架构示意图之一，图3C是图3A所示实施例所适用的检测心电信号的架构示意图之二；如图3A所示，包括如下步骤：
步骤301，对设定时间长度的心电信号进行分割，得到第一设定个数的单心拍。
步骤301的描述可以参见上述图1A所示实施例的描述，在此不再详述。
步骤302，将第一设定个数的单心拍的数据顺次输入到第一卷积神经网络，以提取第一设定个数的单心拍中每一个单心拍对应的特征数据。
步骤302的描述可以参见上述图2A所示实施例的描述，在此不再详述。
步骤303，将第一设定个数的特征数据输入第二卷积神经网络的设定卷积层，将设定时间长度的心电信号输入第二卷积神经网络的输入层，以通过第二卷积神经网络确定该心电信号的病理类别。
步骤303的描述可以参见上述图2A所示实施例的描述，在此不再详述。
下面结合如图3B和图3C对步骤303进行示例性说明。如图3B所示，可以将分割得到的第一设定个数的单心拍的数据顺次输入到第一卷积神经网络，将未分割的原始的心电信号输入到第二卷积神经网络。例如，原始连续的心电信号分割为20个单心拍，则20个单心拍的数据被顺次输入到第一卷积神经网络。其中，该20个单心拍的数据在进入第一卷积神经网络并进行相应处理后，从第一卷积神经网络的设定卷积层输出的特征数据可被缓存起来。例如，从第一卷积神经网络的第二层卷积层输出的20组特征数据可被缓存起来，其中，若每一个单心拍从第二层卷积层输出的特征数据量表示为21*32，则20个单心拍对应的特征数据量可以表示为21*32*20＝420*32。
原始连续的心电信号被输入到第二卷积神经网络的输入层。
在一实施例中，可以将第一卷积神经网络的设定卷积层得到的特征数据注入到第二 卷积神经网络的设定卷积层得到的特征数据。如图3C所示，将第一卷积神经网络的第二层卷积层得到的特征数据，注入到第二卷积神经网络的第二层卷积层得到的特征数据中，将这两组特征数据一起输入到第二卷积神经网络的第三层卷积层。例如，从第二卷积神经网络的第二卷积层得到的特征数据量表示为531*32，则第三层卷积层输入的特征数据量表示为420*32+531*32＝(531-111)*32+531*32＝531*64-111*32，若第三层卷积层能够处理的特征数据量表示为531*64，因此对于第三层卷积层而言，缺少111*32个特征数据，对于缺少的部分，可以通过补零的方式确保第三层卷积层实际输入的特征数据与第三层卷积层需要处理的特征数据相一致。
需要说明的是，图3C仅为示例性说明，可以从第一卷积神经网络的设定卷积层输出不同维度的特征数据，注入到第二卷积神经网络的设定卷积层，从而可灵活调整第二卷积神经网络对第一卷积神经网络的特征数据的依赖程度。这里的设定卷积层可以为某一个卷积层也可以为多个卷积层。例如，可以将从第一卷积神经网络的第二卷积层输出的特征数据，注入到第二卷积神经网络的第二卷积层，同时，还可以将第一卷积神经网络的第三卷积层输出的特征数据注入到第二卷积神经网络的第三卷积层。
本实施例中，通过将第一卷积神经网络的设定卷积层输出的特征数据输入第二卷积神经网络的设定卷积层，可以使第二卷积神经网络的识别结果依赖第一卷积神经网络的特征数据，从而可增强对心电信号的识别性能。
图4A是本发明再一示例性实施例的心电信号的检测方法的流程示意图，图4B是图4A所示实施例所适用的检测心电信号的架构示意图；如图4A所示，包括如下步骤：
步骤401，将设定时间长度的心电信号输入到第二卷积神经网络并通过第二卷积神经网络确定该心电信号的病理类别。
步骤402，若病理类别表示心电信号出现异常，对设定时间长度的心电信号进行分割，得到第一设定个数的单心拍。
步骤403，将得到的第一设定个数的单心拍的数据输入到第一卷积神经网络，并通过第一卷积神经网络确定第一设定个数的单心拍中出现异常的心拍位置。
可选的，在上述步骤403中，可以将第一设定个数的单心拍的数据输入第一卷积神经网络的输入层；通过第一卷积神经网络对第一设定个数的单心拍中的每一个单心拍进行判决，得到判决结果；根据判决结果确定第一设定个数的单心拍中出现异常的心拍位置。
如图4B所示，将一段设定时间长度(例如，30秒)的连续心电信号的数据直接输入第二卷积神经网络，通过对该连续心电信号的数据进行识别，得到病理类别。若病理类别表示心电信号出现异常，可以对该心电信号进行分割，得到例如25个单心拍的数据。这25个单心拍的数据被顺次输入到第一卷积神经网络以被识别，得到判决结果。判决结果可以为0和1组成的序列，例如，0表示异常，1表示正常，则25个单心拍可以对应长度为25位(bit)的0和1的组合，通过识别0所在的位置，即可确定这25个单心拍中出现异常的心拍位置。
本实施例中，在通过第二卷积神经网络识别出的病理类别表示心电信号出现异常时，通过第一卷积神经网络对该心电信号的单心拍逐个进行识别，从而识别出出现异常的心拍位置。
通过上述图2B、图3A和图3B以及图4B所示的架构图可知，本申请中的第一卷积神经网络可视为单心拍识别网络，第二卷积神经网络可视为心电信号检测网络，本申请中第一卷积神经网络和第二卷积神经网络的架构设计，可具有如下有益技术效果：
1)可以对第一卷积神经网络单独训练，降低对整个网络架构的训练难度；
2)易于获取到充足的单心拍数据作为第一卷积神经网络的训练样本，因此第一卷积神经网络可以得到充分的训练，单心拍识别的稳定性和可靠性能够得到保障；
3)可以基于第一卷积神经网络得到的特征数据来增强第二卷积神经网络的训练，从而可解决长序列的连续心拍的心电信号的数据不足的问题；
4)由于心电信号的数量小，因此本申请中的架构设计可适用到嵌入式开发应用中。
图5是本发明一示例性实施例的心电信号的检测装置的结构示意图；如图5所示，该心电信号的检测装置可以包括：第一分割模块51、第一确定模块52、第二确定模块53。其中：
第一分割模块51，用于对设定时间长度的心电信号进行分割，得到第一设定个数的单心拍；
第一确定模块52，用于确定第一分割模块51得到的第一设定个数的单心拍中每一个单心拍对应的特征数据，得到第一设定个数的特征数据；
第二确定模块53，用于基于设定时间长度的心电信号和第一确定模块52确定的第一设定个数的特征数据，确定设定时间长度的心电信号的病理类别。
图6是本发明另一示例性实施例的心电信号的检测装置的结构示意图，如图6所示，在上述图5所示实施例的基础上，第一确定模块52可包括：
第一输入单元521，用于将第一设定个数的单心拍的数据顺次输入到第一卷积神经网络；
提取单元522，用于通过第一卷积神经网络提取第一设定个数的单心拍中每一个单心拍对应的特征数据。
在一实施例中，心电信号的检测装置还包括：
判决模块54，用于通过第一卷积神经网络对第一分割模块51得到的第一设定个数的单心拍中的每一个单心拍进行判决，得到判决结果；
第三确定模块55，用于根据判决模块54得到的判决结果确定第一设定个数的单心拍中出现异常的心拍位置。
在一实施例中，第二确定模块53可包括：
第一确定单元531，用于确定第一设定个数的单心拍中每一个单心拍对应的时序数据，得到第一设定个数的时序数据；
第二输入单元532，用于将第一确定单元531得到的第一设定个数的时序数据和第一设定个数的特征数据输入第二卷积神经网络的输入层；
第二确定单元533，用于通过第二卷积神经网络确定心电信号的病理类别。
其中，第一确定单元531具体可用于：
针对设定时间长度的心电信号中的每一个单心拍，确定该单心拍的R波对应的时间点；
确定与该单心拍的R波前后相邻第二设定个数的单心拍各自的R波对应的时间点；
基于该单心拍的R波对应的时间点以及与该单心拍前后相邻第二设定个数的单心拍各自的R波对应的时间点，确定该单心拍对应的时序数据。
在一实施例中，第二确定模块53包括：
第三输入单元534，用于将第一设定个数的特征数据输入第二卷积神经网络的设定卷积层；
第四输入单元535，用于将设定时间长度的心电信号输入第二卷积神经网络的输 入层；
第三确定单元536，用于通过第二卷积神经网络对第三输入单元534输入的第一设定个数的特征数据和第四输入单元535输入的心电信号进行识别，确定心电信号的病理类别。
图7是本发明又一示例性实施例的心电信号的检测装置的结构示意图，如图7所示，心电信号的检测装置可包括：第四确定模块71、第二分割模块72、第五确定模块73；其中：
第四确定模块71，用于通过第二卷积神经网络确定设定时间长度的心电信号的病理类别；
第二分割模块72，用于若第四确定模块71确定的病理类别表示心电信号出现异常，对设定时间长度的心电信号进行分割，得到第一设定个数的单心拍；
第五确定模块73，用于将第二分割模块72得到的第一设定个数的单心拍的数据输入到第一卷积神经网络，以通过第一卷积神经网络确定第一设定个数的单心拍中出现异常的心拍位置。
在一实施例中，第五确定模块73可包括：
第五输入单元731，用于将第一设定个数的单心拍的数据输入第一卷积神经网络的输入层；
判决单元732，用于通过第五输入单元731输入到第一卷积神经网络对第一设定个数的单心拍中的每一个单心拍进行判决，得到判决结果；
第四确定单元733，用于根据判决单元732得到的判决结果确定第一设定个数的单心拍中出现异常的心拍位置。
本申请内存检测装置的实施例可以应用在电子设备上。装置实施例可以通过软件实现，也可以通过硬件或者软硬件结合的方式实现。以软件实现为例，作为一个逻辑意义上的装置，是通过其所在电子设备的处理器将非易失性存储介质中对应的机器可执行指令读取到内存中运行形成的。从硬件层面而言，如图8所示，为本申请心电信号的检测装置所在电子设备的一种硬件结构图，除了图8所示的处理器801、内存802、网络接口803、以及非易失性存储介质804之外，实施例中装置所在的电子设备通常根据该电子设备的实际功能，还可以包括其他硬件，对此不再赘述。
本领域技术人员在考虑说明书及实践这里公开的发明后，将容易想到本申请的其它实施方案。本申请旨在涵盖本申请的任何变型、用途或者适应性变化，这些变型、用途或者适应性变化遵循本申请的一般性原理并包括本申请未公开的本技术领域中的公知常识或惯用技术手段。说明书和实施例仅被视为示例性的，本申请的真正范围和精神由下面的权利要求指出。
还需要说明的是，术语“包括”、“包含”或者其任何其他变体意在涵盖非排他性的包含，从而使得包括一系列要素的过程、方法、商品或者设备不仅包括那些要素，而且还包括没有明确列出的其他要素，或者是还包括为这种过程、方法、商品或者设备所固有的要素。在没有更多限制的情况下，由语句“包括一个……”限定的要素，并不排除在包括所述要素的过程、方法、商品或者设备中还存在另外的相同要素。
以上所述仅为本申请的较佳实施例而已，并不用以限制本申请，凡在本申请的精神和原则之内，所做的任何修改、等同替换、改进等，均应包含在本申请保护的范围之内。

Claims (17)

  1. 一种心电信号的检测方法，其特征在于，所述方法包括：
    对设定时间长度的心电信号进行分割，得到第一设定个数的单心拍；
    确定所述第一设定个数的单心拍中每一个单心拍对应的特征数据，得到第一设定个数的特征数据；
    基于所述设定时间长度的心电信号和所述第一设定个数的特征数据，确定所述设定时间长度的心电信号的病理类别。
  2. 根据权利要求1所述的方法，其特征在于，确定所述第一设定个数的单心拍中每一个单心拍对应的特征数据，包括：
    将所述第一设定个数的单心拍的数据顺次输入到第一卷积神经网络；
    通过所述第一卷积神经网络提取所述第一设定个数的单心拍中每一个单心拍对应的特征数据。
  3. 根据权利要求2所述的方法，其特征在于，所述方法还包括：
    通过所述第一卷积神经网络对所述第一设定个数的单心拍中的每一个单心拍进行判决，得到判决结果；
    根据所述判决结果确定所述第一设定个数的单心拍中出现异常的心拍位置。
  4. 根据权利要求1-3任一所述的方法，其特征在于，基于所述设定时间长度的心电信号和所述第一设定个数的特征数据，确定所述设定时间长度的心电信号的病理类别，包括：
    确定所述第一设定个数的单心拍中每一个单心拍对应的时序数据，得到第一设定个数的时序数据；
    将所述第一设定个数的时序数据和所述第一设定个数的特征数据输入第二卷积神经网络的输入层，以通过所述第二卷积神经网络确定所述心电信号的病理类别。
  5. 根据权利要求4所述的方法，其特征在于，确定所述单心拍对应的时序数据，包括：
    确定该单心拍的R波对应的时间点；
    确定与该单心拍的R波前后相邻第二设定个数的单心拍各自的R波对应的时间点；
    基于该单心拍的R波对应的时间点以及与该单心拍前后相邻第二设定个数的单心拍各自的R波对应的时间点，确定该单心拍对应的时序数据。
  6. 根据权利要求1所述的方法，其特征在于，基于所述设定时间长度的心电信号和所述第一设定个数的特征数据，确定所述设定时间长度的心电信号的病理类别，包括：
    将所述第一设定个数的特征数据输入第二卷积神经网络的设定卷积层；
    将所述设定时间长度的心电信号输入所述第二卷积神经网络的输入层；
    通过所述第二卷积神经网络确定所述心电信号的病理类别。
  7. 一种心电信号的检测方法，其特征在于，所述方法包括：
    通过第二卷积神经网络确定设定时间长度的心电信号的病理类别；
    若所述病理类别表示所述心电信号出现异常，对所述设定时间长度的心电信号进行分割，得到第一设定个数的单心拍；
    将所述第一设定个数的单心拍的数据输入到第一卷积神经网络，以通过所述第一卷积神经网络确定所述第一设定个数的单心拍中出现异常的心拍位置。
  8. 根据权利要求7所述的方法，其特征在于，通过所述第一卷积神经网络确定所述第一设定个数的单心拍中出现异常的心拍位置，包括：
    将所述第一设定个数的单心拍的数据输入所述第一卷积神经网络的输入层，以通过所述第一卷积神经网络对所述第一设定个数的单心拍中的每一个单心拍进行判决，得到判决结果；
    根据所述判决结果确定所述第一设定个数的单心拍中出现异常的心拍位置。
  9. 一种心电信号的检测装置，其特征在于，所述装置包括：
    第一分割模块，用于对设定时间长度的心电信号进行分割，得到第一设定个数的单心拍；
    第一确定模块，用于确定所述第一分割模块得到的所述第一设定个数的单心拍中每一个单心拍对应的特征数据，得到第一设定个数的特征数据；
    第二确定模块，用于基于所述设定时间长度的心电信号和所述第一确定模块确定的所述第一设定个数的特征数据，确定所述设定时间长度的心电信号的病理类别。
  10. 根据权利要求9所述的装置，其特征在于，所述第一确定模块包括：
    第一输入单元，用于将所述第一设定个数的单心拍的数据顺次输入到第一卷积神经网络；
    提取单元，用于通过所述第一卷积神经网络提取所述第一设定个数的单心拍中每一个单心拍对应的特征数据。
  11. 根据权利要求10所述的装置，其特征在于，所述装置还包括：
    判决模块，用于通过所述第一卷积神经网络对所述第一设定个数的单心拍中的每一个单心拍进行判决，得到判决结果；
    第三确定模块，用于根据所述判决模块得到的所述判决结果确定所述第一设定个数 的单心拍中出现异常的心拍位置。
  12. 根据权利要求9-10任一所述的装置，其特征在于，所述第二确定模块包括：
    第一确定单元，用于确定所述第一设定个数的单心拍中每一个单心拍对应的时序数据，得到第一设定个数的时序数据；
    第二输入单元，用于将所述第一确定单元得到的所述第一设定个数的时序数据和所述第一设定个数的特征数据输入第二卷积神经网络的输入层；
    第二确定单元，用于通过所述第二卷积神经网络确定所述心电信号的病理类别。
  13. 根据权利要求9所述的装置，其特征在于，所述第二确定模块包括：
    第三输入单元，用于将所述第一设定个数的特征数据输入第二卷积神经网络的设定卷积层；
    第四输入单元，用于将所述设定时间长度的心电信号输入所述第二卷积神经网络的输入层；
    第三确定单元，用于通过所述第二卷积神经网络对所述第三输入单元输入的所述第一设定个数的特征数据和所述第四输入单元输入的所述心电信号进行识别，确定所述心电信号的病理类别。
  14. 一种心电信号的检测装置，其特征在于，所述装置包括：
    第四确定模块，用于通过第二卷积神经网络确定设定时间长度的心电信号的病理类别；
    第二分割模块，用于若所述第四确定模块确定的所述病理类别表示所述心电信号出现异常，对所述设定时间长度的心电信号进行分割，得到第一设定个数的单心拍；
    第五确定模块，用于将所述第二分割模块得到的所述第一设定个数的单心拍的数据输入到第一卷积神经网络，以通过所述第一卷积神经网络确定所述第一设定个数的单心拍中出现异常的心拍位置。
  15. 根据权利要求14所述的装置，其特征在于，所述第五确定模块包括：
    第五输入单元，用于将所述第一设定个数的单心拍的数据输入所述第一卷积神经网络的输入层；
    判决单元，用于通过所述第五输入单元输入到所述第一卷积神经网络对所述第一设定个数的单心拍中的每一个单心拍进行判决，得到判决结果；
    第四确定单元，用于根据所述判决单元得到的所述判决结果确定所述第一设定个数的单心拍中出现异常的心拍位置。
  16. 一种机器可读存储介质，其特征在于，所述存储介质存储有机器可执行指令， 所述机器可执行指令用于执行上述权利要求1-8任一所述的心电信号的检测方法。
  17. 一种电子设备，其特征在于，所述电子设备包括：
    处理器；用于存储所述处理器可执行指令的存储介质；
    其中，所述处理器，用于执行上述权利要求1-8任一所述的心电信号的检测方法。
PCT/CN2018/097917 2017-08-03 2018-08-01 心电信号的检测 WO2019024861A1 (zh)

Priority Applications (5)

Application Number Priority Date Filing Date Title
KR1020207005872A KR102451795B1 (ko) 2017-08-03 2018-08-01 심전도 신호의 검출 방법
EP18841762.0A EP3644220A4 (en) 2017-08-03 2018-08-01 ELECTROCARDIOGRAM SIGNAL DETECTION
JP2020526667A JP7065185B2 (ja) 2017-08-03 2018-08-01 心電信号の検出
CA3071699A CA3071699C (en) 2017-08-03 2018-08-01 Detection of electrocardiographic signal
US16/750,568 US11534097B2 (en) 2017-08-03 2020-01-23 Detection of electrocardiographic signal

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201710656849.8A CN107516075B (zh) 2017-08-03 2017-08-03 心电信号的检测方法、装置及电子设备
CN201710656849.8 2017-08-03

Related Child Applications (1)

Application Number Title Priority Date Filing Date
US16/750,568 Continuation US11534097B2 (en) 2017-08-03 2020-01-23 Detection of electrocardiographic signal

Publications (1)

Publication Number Publication Date
WO2019024861A1 true WO2019024861A1 (zh) 2019-02-07

Family

ID=60722226

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2018/097917 WO2019024861A1 (zh) 2017-08-03 2018-08-01 心电信号的检测

Country Status (7)

Country Link
US (1) US11534097B2 (zh)
EP (1) EP3644220A4 (zh)
JP (1) JP7065185B2 (zh)
KR (1) KR102451795B1 (zh)
CN (1) CN107516075B (zh)
CA (1) CA3071699C (zh)
WO (1) WO2019024861A1 (zh)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2020221632A1 (en) * 2019-04-30 2020-11-05 Medicalgorithmics Sa Electrocardiogram signal segmentation
JP6865329B1 (ja) * 2019-07-29 2021-04-28 株式会社カルディオインテリジェンス 表示装置、表示方法及びプログラム
US11521743B2 (en) * 2019-10-21 2022-12-06 Tencent America LLC Framework for performing electrocardiography analysis
US11534097B2 (en) 2017-08-03 2022-12-27 Anhui Huami Information Technology Co., Ltd. Detection of electrocardiographic signal

Families Citing this family (15)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108030488A (zh) * 2017-11-30 2018-05-15 北京医拍智能科技有限公司 基于卷积神经网络的心律失常的检测系统
CN108470158B (zh) * 2018-03-08 2020-05-12 华南理工大学 一种为动态ecg数据寻找误差最小网络计算结构的方法
KR20190114694A (ko) * 2018-03-30 2019-10-10 삼성에스디에스 주식회사 인공지능을 이용한 시계열 데이터 학습 및 분석 방법
EP3847958A4 (en) * 2018-09-04 2021-09-08 Shenzhen Institutes of Advanced Technology ARRHYTHMIA DETECTION METHOD AND APPARATUS, ELECTRONIC DEVICE AND COMPUTER INFORMATION MEDIA
TWI680746B (zh) * 2018-11-21 2020-01-01 晶神醫創股份有限公司 神經事件偵測方法與系統
CN109480824B (zh) * 2018-12-11 2021-10-08 武汉中旗生物医疗电子有限公司 心电波形数据的处理方法、装置和服务器
CN109875579A (zh) * 2019-02-28 2019-06-14 京东方科技集团股份有限公司 情绪健康管理系统和情绪健康管理方法
CN111488793B (zh) * 2020-02-21 2024-04-30 广州视源电子科技股份有限公司 心电信号分类方法、装置、电子设备和存储介质
KR102241799B1 (ko) * 2020-08-06 2021-04-19 주식회사 에이티센스 심전도 신호의 분류 데이터를 제공하는 방법 및 전자 장치
CN112690802B (zh) * 2020-12-25 2023-03-03 平安科技（深圳）有限公司 一种检测心电信号的方法、装置、终端及存储介质
CN112883803B (zh) * 2021-01-20 2023-09-01 武汉中旗生物医疗电子有限公司 一种基于深度学习的心电信号分类方法、装置及存储介质
KR102653258B1 (ko) * 2022-02-18 2024-04-01 주식회사 뷰노 심전도 기술 및 결과 통합 방법
WO2024019523A1 (ko) * 2022-07-22 2024-01-25 주식회사 메디컬에이아이 심전도를 이용한 건강 상태의 예측 방법, 프로그램 및 장치
CN116530996B (zh) * 2023-06-26 2023-08-29 广东技术师范大学 一种面向低测量负荷心电数据的异常预警方法及系统
CN116548980B (zh) * 2023-07-07 2023-09-01 广东技术师范大学 基于脉冲神经网络的长时程心电分类方法及系统

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20080285819A1 (en) * 2006-08-30 2008-11-20 The Trustees Of Columbia University In The City Of New York Systems and method for composite elastography and wave imaging
CN102779234A (zh) * 2012-06-28 2012-11-14 中国科学院苏州纳米技术与纳米仿生研究所 一种心电图分类处理方法及装置
CN104783782A (zh) * 2015-04-13 2015-07-22 深圳市飞马与星月科技研究有限公司 心电信号自动检测方法及装置
CN106344004A (zh) * 2016-09-28 2017-01-25 清华大学 心电信号特征点检测方法及装置
CN106779051A (zh) * 2016-11-24 2017-05-31 厦门中控生物识别信息技术有限公司 一种卷积神经网络模型参数处理方法及系统
CN107516075A (zh) * 2017-08-03 2017-12-26 安徽华米信息科技有限公司 心电信号的检测方法、装置及电子设备

Family Cites Families (11)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
IL155955A0 (en) * 2003-05-15 2003-12-23 Widemed Ltd Adaptive prediction of changes of physiological/pathological states using processing of biomedical signal
TW200927065A (en) 2007-12-28 2009-07-01 Tatung Co Monitoring and control system for cardiopulmonary function and device thereof
EP2523534B1 (en) 2011-05-12 2019-08-07 Ledengin, Inc. Apparatus and methods for tuning of emitter with multiple LEDs to a single color bin
KR20140063100A (ko) * 2012-11-16 2014-05-27 삼성전자주식회사 원격 심질환 관리 장치 및 방법
US20150112157A1 (en) * 2013-10-23 2015-04-23 Quanttus, Inc. Arrhythmia detection
PT3367897T (pt) * 2015-10-27 2021-05-25 Cardiologs Tech Processo automático para traçar ou categorizar um eletrocardiograma
CN105232032A (zh) 2015-11-05 2016-01-13 福州大学 基于小波分析远程心电监护与预警系统及方法
EP3379998A4 (en) 2015-11-23 2019-07-31 Mayo Foundation for Medical Education and Research PROCESSING PHYSIOLOGICAL ELECTRICAL DATA FOR ANALYTICAL ASSESSMENTS
JP6610278B2 (ja) 2016-01-18 2019-11-27 富士通株式会社 機械学習装置、機械学習方法及び機械学習プログラム
US20180032689A1 (en) * 2016-07-29 2018-02-01 Qatar University Method and apparatus for performing feature classification on electrocardiogram data
CN106934799B (zh) * 2017-02-24 2019-09-03 安翰科技(武汉)股份有限公司 胶囊内窥镜图像辅助阅片系统及方法

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20080285819A1 (en) * 2006-08-30 2008-11-20 The Trustees Of Columbia University In The City Of New York Systems and method for composite elastography and wave imaging
CN102779234A (zh) * 2012-06-28 2012-11-14 中国科学院苏州纳米技术与纳米仿生研究所 一种心电图分类处理方法及装置
CN104783782A (zh) * 2015-04-13 2015-07-22 深圳市飞马与星月科技研究有限公司 心电信号自动检测方法及装置
CN106344004A (zh) * 2016-09-28 2017-01-25 清华大学 心电信号特征点检测方法及装置
CN106779051A (zh) * 2016-11-24 2017-05-31 厦门中控生物识别信息技术有限公司 一种卷积神经网络模型参数处理方法及系统
CN107516075A (zh) * 2017-08-03 2017-12-26 安徽华米信息科技有限公司 心电信号的检测方法、装置及电子设备

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
See also references of EP3644220A4 *

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US11534097B2 (en) 2017-08-03 2022-12-27 Anhui Huami Information Technology Co., Ltd. Detection of electrocardiographic signal
WO2020221632A1 (en) * 2019-04-30 2020-11-05 Medicalgorithmics Sa Electrocardiogram signal segmentation
JP6865329B1 (ja) * 2019-07-29 2021-04-28 株式会社カルディオインテリジェンス 表示装置、表示方法及びプログラム
US11521743B2 (en) * 2019-10-21 2022-12-06 Tencent America LLC Framework for performing electrocardiography analysis

Also Published As

Publication number Publication date
US20200155024A1 (en) 2020-05-21
JP2020528336A (ja) 2020-09-24
CN107516075A (zh) 2017-12-26
CA3071699A1 (en) 2019-02-07
KR102451795B1 (ko) 2022-10-06
CN107516075B (zh) 2020-10-09
EP3644220A1 (en) 2020-04-29
EP3644220A4 (en) 2020-11-04
US11534097B2 (en) 2022-12-27
KR20200036908A (ko) 2020-04-07
CA3071699C (en) 2023-02-28
JP7065185B2 (ja) 2022-05-11

Similar Documents

Publication Publication Date Title
WO2019024861A1 (zh) 心电信号的检测
US10869610B2 (en) System and method for identifying cardiac arrhythmias with deep neural networks
Clifford et al. Recent advances in heart sound analysis
Tschannen et al. Heart sound classification using deep structured features
Limam et al. Atrial fibrillation detection and ECG classification based on convolutional recurrent neural network
WO2021017313A1 (zh) 房颤的检测方法、装置、计算机设备及存储介质
CN108511055B (zh) 基于分类器融合及诊断规则的室性早搏识别系统及方法
US11571162B2 (en) Recurrent neural network architecture based classification of atrial fibrillation using single lead ECG
US10271741B2 (en) Biometric scale
CN105726018A (zh) 一种与rr间期无关的房颤自动检测方法
Zhang et al. An ECG heartbeat classification method based on deep convolutional neural network
CN111657925A (zh) 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
CN111956214A (zh) 基于U-net端对端神经网络模型的QRS波自动检测方法
CN110558975A (zh) 一种基于lrf-elm和blstm的心电信号分类方法及系统
CN111053552B (zh) 一种基于深度学习的qrs波检测方法
Ibanez et al. Deep learning for pulmonary nodule ct image retrieval—an online assistance system for novice radiologists
WO2022134472A1 (zh) 一种检测心电信号的方法、装置、终端及存储介质
Pires et al. Automatic diabetic retinopathy detection using BossaNova representation
Golande et al. Optical electrocardiogram based heart disease prediction using hybrid deep learning
Saini et al. Detection of QRS-complex using K-nearest neighbour algorithm
Sadeghi et al. Diagnosing left bundle branch block in 12-lead electrocardiogram using self-attention convolutional neural networks
Sali et al. Software System for Arrhythmia Classification in Real Time of Segmented ECG Signals
Muthuvel et al. Research Article Hybrid Features and Classifier for Classification of ECG Signal
Angelina et al. Classification of Skin Cancer Using ResNet and VGG Deep Learning Network
Lin et al. A Hybrid GCN-LSTM Model for Ventricular Arrhythmia Classification Based on ECG Pattern Similarity

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: 18841762

Country of ref document: EP

Kind code of ref document: A1

ENP Entry into the national phase

Ref document number: 2020526667

Country of ref document: JP

Kind code of ref document: A

ENP Entry into the national phase

Ref document number: 3071699

Country of ref document: CA

NENP Non-entry into the national phase

Ref country code: DE

ENP Entry into the national phase

Ref document number: 2018841762

Country of ref document: EP

Effective date: 20200123

ENP Entry into the national phase

Ref document number: 20207005872

Country of ref document: KR

Kind code of ref document: A