CN111329445A - Atrial fibrillation identification method based on group convolution residual error network and long-term and short-term memory network - Google Patents

Atrial fibrillation identification method based on group convolution residual error network and long-term and short-term memory network Download PDF

Info

Publication number
CN111329445A
CN111329445A CN202010106864.7A CN202010106864A CN111329445A CN 111329445 A CN111329445 A CN 111329445A CN 202010106864 A CN202010106864 A CN 202010106864A CN 111329445 A CN111329445 A CN 111329445A
Authority
CN
China
Prior art keywords
layer
network
output
group
convolution
Prior art date
Legal status (The legal status 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 status listed.)
Granted
Application number
CN202010106864.7A
Other languages
Chinese (zh)
Other versions
CN111329445B (en
Inventor
余锭能
吕俊
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Guangdong University of Technology
Original Assignee
Guangdong University of Technology
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 Guangdong University of Technology filed Critical Guangdong University of Technology
Priority to CN202010106864.7A priority Critical patent/CN111329445B/en
Publication of CN111329445A publication Critical patent/CN111329445A/en
Application granted granted Critical
Publication of CN111329445B publication Critical patent/CN111329445B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

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
    • A61B5/72—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
    • 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
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7235—Details of waveform analysis
    • A61B5/725—Details of waveform analysis using specific filters therefor, e.g. Kalman or adaptive filters
    • 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
    • A61B5/7285—Specific aspects of physiological measurement analysis for synchronising or triggering a physiological measurement or image acquisition with a physiological event or waveform, e.g. an ECG signal
    • 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
    • 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
    • G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/08—Feature extraction
    • 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

Landscapes

  • Health & Medical Sciences (AREA)
  • Engineering & Computer Science (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Physics & Mathematics (AREA)
  • Artificial Intelligence (AREA)
  • Public Health (AREA)
  • Biomedical Technology (AREA)
  • Veterinary Medicine (AREA)
  • General Health & Medical Sciences (AREA)
  • Animal Behavior & Ethology (AREA)
  • Surgery (AREA)
  • Molecular Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Medical Informatics (AREA)
  • Heart & Thoracic Surgery (AREA)
  • Biophysics (AREA)
  • Pathology (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Cardiology (AREA)
  • Psychiatry (AREA)
  • Evolutionary Computation (AREA)
  • Physiology (AREA)
  • Signal Processing (AREA)
  • General Engineering & Computer Science (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • General Physics & Mathematics (AREA)
  • Evolutionary Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Fuzzy Systems (AREA)
  • Mathematical Physics (AREA)
  • Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)

Abstract

The invention discloses an atrial fibrillation identification method based on a group convolution residual error network and a long-short term memory network, and relates to the field of atrial fibrillation machine identification. According to the invention, on the basis of a network structure of a group convolution residual error network and a long-short term memory network, three channels are respectively used for extracting features of electrocardiosignals of three different frequency bands, then LSTM is used for carrying out feature analysis on a time domain, and finally, the electrocardiosignal segments are classified into normal segments, atrial fibrillation segments, segments with larger noise and other beat segments.

Description

Atrial fibrillation identification method based on group convolution residual error network and long-term and short-term memory network
Technical Field
The invention relates to the technical field of atrial fibrillation machine identification, in particular to an atrial fibrillation identification method based on a group convolution residual error network and a long-term and short-term memory network.
Background
Atrial fibrillation (atrial fibrillation) is the most common arrhythmia disease, with prevalence rates around 0.4-1% in the general population, increasing to 8% in people over 80 years of age. The occurrence of atrial fibrillation symptoms is closely related to coronary heart disease, hypertension, heart failure and other diseases. Atrial fibrillation does not itself directly threaten the life and health of the patient. Without timely treatment, however, atrial fibrillation can cause serious complications, such as heart failure and stroke. Heart failure can seriously affect the quality of life of patients and stroke is listed as the second leading cause of death in the world by the world health organization, and both seriously threaten the life and health of people. Therefore, early detection of atrial fibrillation is critical to prevent its induced disorders.
The prior art generally adopts the following three methods for identifying atrial fibrillation:
(1) ruhi Mahajan et al propose a feature extraction method combining probability symbol pattern recognition and sample entropy, and the morphological change of electrocardiosignals is represented by the method so as to recognize atrial fibrillation signals;
(2) xiaoyan Xu et al propose a framework combining Modified Frequency Slice Wavelet Transform (MFSWT) and convolutional neural network, and automatically recognize atrial fibrillation ecg signals under the framework;
(3) mohamed Limam et al propose a convolutional-based recurrent neural network, which includes two independent neural networks, extracts relevant patterns from ECG and heart rate, respectively, then fuses the patterns into one recurrent neural network, takes charge of extracting the sequence of patterns through the recurrent neural network, and then evaluates the final decision through a support vector machine.
The method for identifying the atrial fibrillation based on the prior art has the defects of overlong detection time and lower real-time performance of monitoring and analyzing the atrial fibrillation. The relatively clean electrocardiosignals are often needed in the monitoring and detecting process, the noisy segments are easily misreported to atrial fibrillation signals, and the identification accuracy is not high.
Disclosure of Invention
In order to solve the defects of the prior art, the invention provides an atrial fibrillation identification method based on a group convolution residual error network and a long-short term memory network.
The technical scheme adopted by the invention for realizing the technical effects is as follows:
the atrial fibrillation identification method based on the group convolution residual error network and the long-short term memory network comprises the following steps:
s1, preprocessing data, namely segmenting the original electrocardiosignals into n segments by using a sliding window with the window length of 2500 and the step length of 1, down-sampling the segmented segments to 250Hz, and then performing low-pass filtering, band-pass filtering and high-pass filtering on the down-sampled signals to obtain three frequency band sequences d1, d2 and d 3;
s2, correspondingly inputting the data of the three frequency band sequences d1, d2 and d3 obtained in the step S1 into three residual error networks one by one for feature extraction;
s3, inputting the characteristic signal extracted in step S2 to LSTM, and extracting the time-series characteristic thereof in a pattern sequence;
s4, tiling the output of the LSTM by using a full connection layer, and then calculating the loss by using a cross entropy function;
and S5, judging whether the loss is less than a threshold value, if so, saving the data model, and otherwise, performing back propagation to continue training the model.
Preferably, in the above method for identifying atrial fibrillation based on the group convolution residual error network and the long-short term memory network, the residual error network is formed by stacking a plurality of residual error modules, and the residual error modules include a group convolution block for performing a group convolution on an original single-layer convolution.
Preferably, in the above atrial fibrillation recognition method based on the group convolution residual error network and the long-term and short-term memory network, the first part of the residual error module is a convolution layer, a batch normalization layer and an activation layer; the second part is a group rolling block, a batch standardization layer and an activation layer; the third part is a layer of convolution, batch normalization layer and activation layer, and the data input into the activation layer of the third part is the sum of the data stream output by batch normalization in the third part and the input data of the residual module.
Preferably, in the above atrial fibrillation recognition method based on the group convolution residual error network and the long-short term memory network, the expression of the first part of the residual error module is as follows:
Fm，n＝ReLU(BN(conv1D(wm，n，1，xm，n)))；
wherein ,
Figure BDA0002388292970000031
a residual network sequence number is indicated,
Figure BDA0002388292970000032
denotes the residual module number, BN denotes batch normalization, conv1D (. cndot.) denotes one-dimensional convolution, wm，n，1Convolution kernel parameter, x, representing the first part of the nth residual block of the mth network channelm，nRepresenting the input of the nth residual block of the mth network channel,
Figure BDA0002388292970000033
ReLU represents the activation function, whose expression is:
Figure BDA0002388292970000034
preferably, in the above atrial fibrillation recognition method based on the group convolution residual error network and the long-short term memory network, the expression of the second part of the residual error module is as follows:
Figure BDA0002388292970000035
wherein x is [ x ]1，x2，...，xg]Representing the input of each group in the group convolution block, g representing the number of groups convolved by the group, i representing the ordinal number, Ti(x) A stack computation process representing the three convolutions in each of the groups of convolution blocks, R (x) represents the output of the group convolution; the group volume block is represented as
Figure BDA0002388292970000036
wiThe weight parameter of each group is represented.
Preferably, in the above group convolution-based residual networkIn the atrial fibrillation identification method of the long-short term memory network, the expression of the third part of the residual error module is as follows: THm，n＝ReLU(BN(conv1D(wm，n，3，Sm，n))+xm，n) wherein ,
Figure BDA0002388292970000041
the expression of the splicing vector after the network channels of the three residual error networks output is as follows: p ═ cat [ TH ]1，18，TH2，18，TH3，18]，
Figure BDA0002388292970000042
The operation represents stitching the vectors.
Preferably, in the above atrial fibrillation recognition method based on the group convolution residual network and the long-short term memory network, the vector p obtained by splicing the outputs of the three network channels is input to a forgetting gate in the LSTM, and the forgetting gate at the time t is obtained as follows: f. oft＝σ(Wf·cat[ht-1，pt]+bf)；
Wherein σ is a sigmiod function, WfWeight of forgetting gate, ht-1For the hidden layer output of the last cell, ptIs an input at time t, bfA bias for a forgetting gate; wherein, Wf＝dc×(dh+dp)，dpTo input dimension, dhTo hide the layer dimension, dcDimension that is the cell state;
the input gates at time t are: i.e. it＝σ(Wi·cat[ht-1，pt]+bi)； wherein ,WiAs the weight of the input gate, biIs the bias of the input gate;
the input cell state at time t is:
Figure BDA0002388292970000043
wherein ,wcAs weights of the states of the input cells, bcA bias that is an input cell state;
the overall unit state at time t is:
Figure BDA0002388292970000044
the output gate at time t is: ot＝σ(Wo·cat[ht-1，pt]+bo； wherein ,woAs weights of output gates, boIs the offset of the output gate;
the overall unit output at time t is: h ist＝ft°tanh(ct)。
Preferably, in the aforementioned atrial fibrillation recognition method based on the group convolution residual error network and the long-short term memory network, in step S4, the output classification layer of the LSTM includes two fully-connected layers and an error calculation layer, each activation function is ReLU, and the output expression of the first fully-connected layer is:
Figure BDA0002388292970000051
the output expression of the second layer full connection is
Figure BDA0002388292970000052
Then, performing error calculation on the output of the last full-connection layer by using a cross entropy function;
where d represents the number of neurons in the fully connected layer, h represents the output vector of LSTM, and y (h) represents the calculation of the weights of h and the fully connected layer.
Preferably, in the above atrial fibrillation recognition method based on the group convolution residual error network and the long-short term memory network, in the data preprocessing process of step S1, the cut-off frequency of the low-pass filtering is 5Hz, the cut-off frequency of the band-pass filtering is 5Hz, 13Hz, and the cut-off frequency of the high-pass filtering is 13 Hz.
The invention has the beneficial effects that: the invention can improve the classification accuracy without increasing the parameter complexity, benefits from the topological structure of the group rolling block in the residual error module, and simultaneously reduces the data volume of the hyper-parameters. On the network structure, three channels respectively extract features of electrocardiosignals of three different frequency bands, and then LSTM is adopted to perform feature analysis on a time domain, and finally electrocardiosignal segments are classified into normal segments, atrial fibrillation segments, segments with larger noise and other beat segments. By adopting the network model, the atrial fibrillation identification correctness under the condition of noisy interference can be improved, the analysis time is shortened, and the real-time performance of the algorithm is improved.
Drawings
FIG. 1 is a schematic flow diagram of the present invention;
FIG. 2 is a diagram of a network architecture of the present invention;
FIG. 3 is a block diagram of a residual network and residual module according to the present invention;
fig. 4 is a block diagram of a group volume block according to the present invention.
Detailed Description
For a further understanding of the invention, reference should be made to the following description taken in conjunction with the accompanying drawings and specific examples.
In the description of the present invention, it is to be understood that the terms "upper", "lower", "front", "rear", "left", "right", "top", "bottom", "inner", "outer", and the like, indicate orientations or positional relationships based on the orientations or positional relationships shown in the drawings, are merely for convenience in describing the present invention and simplifying the description, and do not indicate or imply that the device or element being referred to must have a particular orientation, be constructed and operated in a particular orientation, and thus, should not be construed as limiting the present invention.
As shown in fig. 1, the method for identifying atrial fibrillation based on a group convolution residual error network and a long-short term memory network according to an embodiment of the present invention includes the following steps:
s1, preprocessing data, namely segmenting the original electrocardiosignals into n segments by using a sliding window with the window length of 2500 and the step length of 1, down-sampling the segmented segments to 250Hz, and then performing low-pass filtering, band-pass filtering and high-pass filtering on the down-sampled signals to obtain three frequency band sequences d1, d2 and d 3;
s2, correspondingly inputting the data of the three frequency band sequences d1, d2 and d3 obtained in the step S1 into three residual error networks one by one for feature extraction;
s3, inputting the characteristic signal extracted in step S2 to LSTM, and extracting the time-series characteristic thereof in a pattern sequence;
s4, tiling the output of the LSTM by using a full connection layer, and then calculating the loss by using a cross entropy function;
and S5, judging whether the loss is less than a threshold value, if so, saving the data model, and otherwise, performing back propagation to continue training the model.
The network structure constructed based on the group convolution residual error network and the long-short term memory network is shown in fig. 2, three frequency band sequences d1, d2 and d3 obtained by processing original electrocardiosignals through data are respectively and correspondingly input into the three residual error networks, feature extraction is carried out through the three residual error networks, then mode sequence extraction is carried out on the time sequence features of the three frequency band sequences through the LSTM, namely the long-short term memory network, and then output of the LSTM is tiled through two full connection layers.
As shown in fig. 3, which is a structural diagram of the residual error network and the residual error module, in the preferred embodiment of the present invention, as shown in the left diagram of fig. 3, the residual error network is formed by stacking a plurality of residual error modules. In the preferred embodiment of the present invention, the number of residual error modules is 18, and each residual error module includes a group convolution block for performing a group convolution on the original single-layer convolution. As shown in the right diagram of fig. 3, the first part of the residual block is composed of convolution with convolution kernel size S and number of channels F, and Batch Normalization (BN) and activation layer (ReLU) and the second part of the residual block is composed of group convolution block, batch normalization layer and activation layer and the third part of the residual block is composed of convolution with convolution kernel size S and number of channels F, and batch normalization layer and activation layer. The data input to the active layer of the third portion is the sum of the data stream of the batch normalized output at the third portion and the input data of the residual module.
As shown in fig. 4, which is a structural diagram of the group of convolution blocks according to the present invention, the original single convolution process is divided into 32 groups for convolution respectively, each group is composed of three convolutions, the sizes of convolution kernels are 1, 3 and 1, and the convolution step size is 1.
Specifically, in a preferred embodiment of the present invention, the expression of the first part of the residual error module is:
Fm，n＝ReLU(BN(conv1D(wm，n，1，xm，n)))；
wherein ,
Figure BDA0002388292970000071
a residual network sequence number is indicated,
Figure BDA0002388292970000072
denotes the residual module number, BN denotes batch normalization, conv1D (. cndot.) denotes one-dimensional convolution, wm，n，1Convolution kernel parameter, x, representing the first part of the nth residual block of the mth network channelm，nRepresenting the input of the nth residual block of the mth network channel,
Figure BDA0002388292970000073
ReLU represents the activation function, whose expression is:
Figure BDA0002388292970000081
the second part of the residual module is a group rolling block, a batch normalization layer and an activation layer, and can be represented as follows:
Figure BDA0002388292970000082
the group volume block can be generally expressed as
Figure BDA0002388292970000083
Wherein x is [ x ]1，x2，...，xg]Representing the input of each group in the group convolution block, g representing the group number of the group convolution, i representing the ordinal number; w is aiThe weight parameter, i.e., the convolution kernel, of each group is represented. Final combination of each group of the group rolling blocks
Figure BDA0002388292970000084
Expressions to group into tiles
Figure BDA0002388292970000085
The transformation is of the general form:
Figure BDA0002388292970000086
wherein ,Ti(x) Represents the stack computation process of the three convolutions in each group of the group convolution block, and r (x) represents the output result of the group convolution. Will be a formula
Figure BDA0002388292970000087
X in (1) is replaced by Fm，nThen a formula can be obtained
Figure BDA0002388292970000088
The expression of the third part of the residual block is: THm，n＝ReLU(BN(conv1D(wm，n，3，Sm，n))+xm，n) (ii) a The expression of the splicing vector after the network channels of the three residual error networks output is as follows: p ═ cat [ TH ]1，18，TH2，18，TH3，18]，
Figure BDA0002388292970000089
The operation represents stitching the vectors.
Specifically, in the preferred embodiment of the present invention, the vector p obtained by splicing the outputs of the three network channels is input to a forgetting gate in the LSTM, and the forgetting gate at time t is obtained as follows: f. oft＝σ(Wf·cat[ht-1，pt]+bf)；
Wherein σ is a sigmiod function, WfWeight of forgetting gate, ht-1For the hidden layer output of the last cell, ptIs an input at time t, bfA bias for a forgetting gate; wherein, Wf＝dc×(dh+dp)，dpTo input dimension, dhTo hide the layer dimension, dcDimension that is the cell state;
the input gates at time t are: i.e. it＝σ(Wi·cat[ht-1，pt]+bi)； wherein ,WiAs the weight of the input gate, biIs the bias of the input gate;
the input cell state at time t is:
Figure BDA0002388292970000091
wherein ,wcAs weights of the states of the input cells, bcA bias that is an input cell state;
the overall unit state at time t is:
Figure BDA0002388292970000094
the output gate at time t is: ot＝σ(Wo·cat[ht-1，pt]+bo)； wherein ,woAs weights of output gates, boIs the offset of the output gate;
the overall unit output at time t is: h ist＝ft°tanh(ct)。
In step S4, the output classification layer of the LSTM includes two fully-connected layers and an error calculation layer, the respective activation functions are ReLU, and the output expression of the first fully-connected layer is:
Figure BDA0002388292970000092
the output expression of the second layer full connection is
Figure BDA0002388292970000093
The cross entropy function is then used to perform an error calculation on the output of the last fully connected layer.
Where d represents the number of neurons in the fully connected layer, h represents the output vector of LSTM, and y (h) represents the calculation of the weights of h and the fully connected layer.
Specifically, in the preferred embodiment of the present invention, during the data preprocessing of step S1, the cut-off frequency of the low-pass filtering is 5Hz, the cut-off frequency of the band-pass filtering is 5Hz, 13Hz, and the cut-off frequency of the high-pass filtering is 13Hz, respectively.
The experiment results show that the electrocardiogram monitor and the electrocardiogram machine implanted into the algorithm model can monitor the electrocardiogram state of a testee in real time, and after the testee generates atrial fibrillation signals, the monitor can accurately alarm about 5 seconds at the fastest speed, so that the real-time performance is effectively improved. Meanwhile, the false alarm rate is effectively reduced when the interference of the electrocardiosignal is large.
In summary, the present invention can improve the classification accuracy without increasing the complexity of the parameters, and reduce the amount of data of the hyper-parameters while benefiting from the topology of the group rolling block in the residual module. On the network structure, three channels respectively extract features of electrocardiosignals of three different frequency bands, and then LSTM is adopted to perform feature analysis on a time domain, and finally electrocardiosignal segments are classified into normal segments, atrial fibrillation segments, segments with larger noise and other beat segments. By adopting the network model, the atrial fibrillation identification correctness under the condition of noisy interference can be improved, the analysis time is shortened, and the real-time performance of the algorithm is improved.
The foregoing shows and describes the general principles, essential features, and advantages of the invention. It will be understood by those skilled in the art that the present invention is not limited to the embodiments described above, which are merely illustrative of the principles of the invention, but that various changes and modifications may be made without departing from the spirit and scope of the invention, which is defined by the appended claims and their equivalents.

Claims (9)

1. The atrial fibrillation identification method based on the group convolution residual error network and the long-short term memory network is characterized by comprising the following steps of:
s1, preprocessing data, namely segmenting the original electrocardiosignals into n segments by using a sliding window with the window length of 2500 and the step length of 1, down-sampling the segmented segments to 250Hz, and then performing low-pass filtering, band-pass filtering and high-pass filtering on the down-sampled signals to obtain three frequency band sequences d1, d2 and d 3;
s2, correspondingly inputting the data of the three frequency band sequences d1, d2 and d3 obtained in the step S1 into three residual error networks one by one for feature extraction;
s3, inputting the characteristic signal extracted in step S2 to LSTM, and extracting the time-series characteristic thereof in a pattern sequence;
s4, tiling the output of the LSTM by using a full connection layer, and then calculating the loss by using a cross entropy function;
and S5, judging whether the loss is less than a threshold value, if so, saving the data model, and otherwise, performing back propagation to continue training the model.
2. The method according to claim 1, wherein the residual network is formed by stacking a plurality of residual modules, and the residual modules comprise a group convolution block for performing a group convolution on the original single-layer convolution.
3. The method for identifying atrial fibrillation according to claim 2, wherein the first part of the residual error module is a convolutional layer, a batch normalization layer and an activation layer; the second part is a group rolling block, a batch standardization layer and an activation layer; the third part is a layer of convolution, batch normalization layer and activation layer, and the data input into the activation layer of the third part is the sum of the data stream output by batch normalization in the third part and the input data of the residual module.
4. The method of claim 3, wherein the expression of the first part of the residual error module is:
Fm，n＝ReLU(BN(conv1D(wm，n，1，xm，n)))；
wherein ,
Figure FDA0002388292960000021
a residual network sequence number is indicated,
Figure FDA0002388292960000022
denotes the residual module number, BN denotes batch normalization, conv1D (. cndot.) denotes one-dimensional convolution, wm，n，1Convolution kernel parameter, x, representing the first part of the nth residual block of the mth network channelm，nRepresenting the input of the nth residual block of the mth network channel,
Figure FDA0002388292960000023
ReLU represents the activation function, whose expression is:
Figure FDA0002388292960000024
5. the method of claim 4, wherein the expression of the second part of the residual error module is:
Figure FDA0002388292960000025
wherein x is [ x ]1，x2，...，xg]Representing the input of each group in the group convolution block, g representing the number of groups convolved by the group, i representing the ordinal number, Ti(x) A stack computation process representing the three convolutions in each of the groups of convolution blocks, R (x) represents the output of the group convolution; the group volume block is represented as
Figure FDA0002388292960000026
wiThe weight parameter of each group is represented.
6. The method of claim 5, wherein the expression of the third part of the residual error module is: THm，n＝ReLU(BU(conv1D(wm，n，3，Sm，n))+xm，n), wherein ,
Figure FDA0002388292960000027
the expression of the splicing vector after the network channels of the three residual error networks output is as follows: p ═ cat [ TH ]1，18，TH2，18，TH3，18]，
Figure FDA0002388292960000028
The operation represents stitching the vectors.
7. The atrial fibrillation recognition method based on the group convolution residual error network and the long-short term memory network as claimed in claim 6, wherein the vector p obtained by splicing the outputs of the three network channels is input to a forgetting gate in the LSTM, and the forgetting gate at the time t is obtained by: f. oft＝σ(Wf·cat[ht-1，pt]+bf)；
Wherein σ is a sigmiod function, WfWeight of forgetting gate, ht-1For the hidden layer output of the last cell, ptIs an input at time t, bfA bias for a forgetting gate; wherein, Wf＝dcx(dh+dp)，dpTo input dimension, dhTo hide the layer dimension, dcDimension that is the cell state;
the input gates at time t are: i.e. it＝σ(Wi·cat[ht-1，pt]+bi)； wherein ,WiAs the weight of the input gate, biIs the bias of the input gate;
the input cell state at time t is:
Figure FDA0002388292960000031
wherein ,wcAs weights of the states of the input cells, bcA bias that is an input cell state;
the overall unit state at time t is:
Figure FDA0002388292960000032
the output gate at time t is: ot＝σ(Wo·cat[ht-1，pt]+bo； wherein ,woAs weights of output gates, boIs the offset of the output gate;
the overall unit output at time t is:
Figure FDA0002388292960000033
8. the method for atrial fibrillation recognition based on the group convolution residual error network and the long-short term memory network, wherein in step S4, the output classification layer of the LSTM includes two fully-connected layers and an error calculation layer, the respective activation functions are ReLU, and the output expression of the first fully-connected layer is:the output expression of the second layer full connection is
Figure FDA0002388292960000035
Then, performing error calculation on the output of the last full-connection layer by using a cross entropy function;
where d represents the number of neurons in the fully connected layer, h represents the output vector of LSTM, and y (h) represents the calculation of the weights of h and the fully connected layer.
9. The method for atrial fibrillation recognition based on the group convolution residual error network and the long-short term memory network of claim 1, wherein during the data preprocessing of step S1, the cut-off frequency of the low-pass filtering is 5Hz, the cut-off frequency of the band-pass filtering is 5Hz and 13Hz respectively, and the cut-off frequency of the high-pass filtering is 13 Hz.
CN202010106864.7A 2020-02-20 2020-02-20 Atrial fibrillation identification method based on group convolution residual error network and long-term and short-term memory network Active CN111329445B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010106864.7A CN111329445B (en) 2020-02-20 2020-02-20 Atrial fibrillation identification method based on group convolution residual error network and long-term and short-term memory network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010106864.7A CN111329445B (en) 2020-02-20 2020-02-20 Atrial fibrillation identification method based on group convolution residual error network and long-term and short-term memory network

Publications (2)

Publication Number Publication Date
CN111329445A true CN111329445A (en) 2020-06-26
CN111329445B CN111329445B (en) 2023-09-15

Family

ID=71173874

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010106864.7A Active CN111329445B (en) 2020-02-20 2020-02-20 Atrial fibrillation identification method based on group convolution residual error network and long-term and short-term memory network

Country Status (1)

Country Link
CN (1) CN111329445B (en)

Cited By (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111920375A (en) * 2020-09-30 2020-11-13 平安科技（深圳）有限公司 Vascular Endothelial Growth Factor (VEGF) resistance curative effect prediction device and method
CN111920405A (en) * 2020-09-15 2020-11-13 齐鲁工业大学 Atrial fibrillation signal identification system and method
CN112244861A (en) * 2020-10-09 2021-01-22 广东工业大学 Single-lead electrocardiosignal f-wave extraction method
CN112426160A (en) * 2020-11-30 2021-03-02 贵州省人民医院 Electrocardiosignal type identification method and device
CN112836569A (en) * 2020-12-15 2021-05-25 泰山学院 Underwater acoustic communication signal identification method, system and equipment based on sequence convolution network
CN113288161A (en) * 2021-06-02 2021-08-24 南京信息工程大学 Single-lead ECG signal classification method and system based on improved residual error network
CN114858467A (en) * 2022-05-26 2022-08-05 上海交通大学 Anti-noise and cross-noise-domain misfire diagnosis method and system for diesel engine

Citations (13)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2006068145A1 (en) * 2004-12-22 2006-06-29 Dainippon Sumitomo Pharma Co., Ltd Electrocardiogram waveform correction display and electrocardiogram waveform correction display method
CN104783780A (en) * 2015-04-13 2015-07-22 深圳市飞马与星月科技研究有限公司 Electrocardiosignal denoising method and device
CN106096579A (en) * 2016-06-22 2016-11-09 天津理工大学 A kind of method of ECG signal processing
CN108113666A (en) * 2017-12-19 2018-06-05 中国科学院深圳先进技术研究院 Recognition methods, device and the equipment of atrial fibrillation signal
CN108606798A (en) * 2018-05-10 2018-10-02 东北大学 Contactless atrial fibrillation intelligent checking system based on depth convolution residual error network
CN109303560A (en) * 2018-11-01 2019-02-05 杭州质子科技有限公司 A kind of atrial fibrillation recognition methods of electrocardiosignal in short-term based on convolution residual error network and transfer learning
CN110046604A (en) * 2019-04-25 2019-07-23 成都信息工程大学 A kind of single lead ECG arrhythmia detection classification method based on residual error network
CN110226918A (en) * 2019-06-26 2019-09-13 广州视源电子科技股份有限公司 Electrocardiosignal type detection method, apparatus, computer equipment and storage medium
CN110309758A (en) * 2019-06-26 2019-10-08 广州视源电子科技股份有限公司 Electrocardiosignal feature extracting method, device, computer equipment and storage medium
CN110327039A (en) * 2019-04-15 2019-10-15 四川大学 Atrial fibrillation classification method based on convolutional neural networks
CN110393519A (en) * 2019-08-19 2019-11-01 广州视源电子科技股份有限公司 Analysis method, device, storage medium and the processor of electrocardiosignal
CN110522440A (en) * 2019-08-12 2019-12-03 广州视源电子科技股份有限公司 Electrocardiosignal identification device based on grouping convolutional neural networks
CN110680278A (en) * 2019-09-10 2020-01-14 广州视源电子科技股份有限公司 Electrocardiosignal recognition device based on convolutional neural network

Patent Citations (13)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2006068145A1 (en) * 2004-12-22 2006-06-29 Dainippon Sumitomo Pharma Co., Ltd Electrocardiogram waveform correction display and electrocardiogram waveform correction display method
CN104783780A (en) * 2015-04-13 2015-07-22 深圳市飞马与星月科技研究有限公司 Electrocardiosignal denoising method and device
CN106096579A (en) * 2016-06-22 2016-11-09 天津理工大学 A kind of method of ECG signal processing
CN108113666A (en) * 2017-12-19 2018-06-05 中国科学院深圳先进技术研究院 Recognition methods, device and the equipment of atrial fibrillation signal
CN108606798A (en) * 2018-05-10 2018-10-02 东北大学 Contactless atrial fibrillation intelligent checking system based on depth convolution residual error network
CN109303560A (en) * 2018-11-01 2019-02-05 杭州质子科技有限公司 A kind of atrial fibrillation recognition methods of electrocardiosignal in short-term based on convolution residual error network and transfer learning
CN110327039A (en) * 2019-04-15 2019-10-15 四川大学 Atrial fibrillation classification method based on convolutional neural networks
CN110046604A (en) * 2019-04-25 2019-07-23 成都信息工程大学 A kind of single lead ECG arrhythmia detection classification method based on residual error network
CN110226918A (en) * 2019-06-26 2019-09-13 广州视源电子科技股份有限公司 Electrocardiosignal type detection method, apparatus, computer equipment and storage medium
CN110309758A (en) * 2019-06-26 2019-10-08 广州视源电子科技股份有限公司 Electrocardiosignal feature extracting method, device, computer equipment and storage medium
CN110522440A (en) * 2019-08-12 2019-12-03 广州视源电子科技股份有限公司 Electrocardiosignal identification device based on grouping convolutional neural networks
CN110393519A (en) * 2019-08-19 2019-11-01 广州视源电子科技股份有限公司 Analysis method, device, storage medium and the processor of electrocardiosignal
CN110680278A (en) * 2019-09-10 2020-01-14 广州视源电子科技股份有限公司 Electrocardiosignal recognition device based on convolutional neural network

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
KENSHO HARA ET AL.: ""Can Spatiotemporal 3D CNNs Retrace the History of 2D CNNs and ImageNet?"", pages 3 *
SAINING XIE ET AL.: "\"Aggregated Residual Transformations for Deep Neural Networks\"", pages 3 *

Cited By (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111920405A (en) * 2020-09-15 2020-11-13 齐鲁工业大学 Atrial fibrillation signal identification system and method
CN111920375A (en) * 2020-09-30 2020-11-13 平安科技（深圳）有限公司 Vascular Endothelial Growth Factor (VEGF) resistance curative effect prediction device and method
CN111920375B (en) * 2020-09-30 2021-01-15 平安科技（深圳）有限公司 Vascular Endothelial Growth Factor (VEGF) resistance curative effect prediction device and method
CN112244861A (en) * 2020-10-09 2021-01-22 广东工业大学 Single-lead electrocardiosignal f-wave extraction method
CN112426160A (en) * 2020-11-30 2021-03-02 贵州省人民医院 Electrocardiosignal type identification method and device
CN112836569A (en) * 2020-12-15 2021-05-25 泰山学院 Underwater acoustic communication signal identification method, system and equipment based on sequence convolution network
CN112836569B (en) * 2020-12-15 2023-01-03 泰山学院 Underwater acoustic communication signal identification method, system and equipment based on sequence convolution network
CN113288161A (en) * 2021-06-02 2021-08-24 南京信息工程大学 Single-lead ECG signal classification method and system based on improved residual error network
CN114858467A (en) * 2022-05-26 2022-08-05 上海交通大学 Anti-noise and cross-noise-domain misfire diagnosis method and system for diesel engine
CN114858467B (en) * 2022-05-26 2023-05-26 上海交通大学 Diesel engine anti-noise and cross-noise domain fire diagnosis method and system

Also Published As

Publication number Publication date
CN111329445B (en) 2023-09-15

Similar Documents

Publication Publication Date Title
CN111329445A (en) Atrial fibrillation identification method based on group convolution residual error network and long-term and short-term memory network
Wu et al. A study on arrhythmia via ECG signal classification using the convolutional neural network
CN111261282A (en) Sepsis early prediction method based on machine learning
US8649854B2 (en) Method and device for the determination of murmur frequency band
Priscila et al. Heart Disease Prediction Using Integer-Coded Genetic Algorithm (ICGA) Based Particle Clonal Neural Network (ICGA-PCNN)
CN110619322A (en) Multi-lead electrocardio abnormal signal identification method and system based on multi-flow convolution cyclic neural network
CN111009324B (en) Auxiliary diagnosis system and method for mild cognitive impairment through multi-feature analysis of brain network
CN111297349A (en) Machine learning-based heart rhythm classification system
CN110638430B (en) Method for building cascade neural network ECG signal arrhythmia classification model
CN113288162A (en) Short-term electrocardiosignal atrial fibrillation automatic detection system based on self-adaptive attention mechanism
CN113080993B (en) Automatic arrhythmia analysis method based on improved multi-base classifier integration
CN110491506A (en) Auricular fibrillation prediction model and its forecasting system
Aksakallı et al. Kidney x-ray images classification using machine learning and deep learning methods
CN113349793A (en) Arrhythmia classification method based on layered attention model
Rujoie et al. Classification and evaluation of the severity of tricuspid regurgitation using phonocardiogram
CN115281688A (en) Cardiac hypertrophy multi-label detection system based on multi-mode deep learning
Prakash et al. A system for automatic cardiac arrhythmia recognition using electrocardiogram signal
CN114847905A (en) Arrhythmia data detection and identification method and system
Praveen et al. Chronic kidney disease prediction using ML-based neuro-fuzzy model
US6941288B2 (en) Online learning method in a decision system
Hilmy et al. Ensembles of convolutional neural networks for skin lesion dermoscopy images classification
Nimitha et al. Supervised chromosomal anomaly detection using VGG-16 CNN model
Chowdhury et al. Combining deep learning with traditional machine learning to improve phonocardiography classification accuracy
Wen et al. ECG beat classification using GreyART network
VK et al. Down syndrome identification and classification using facial features with neural network

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant