CN114190889A - Electrocardiosignal classification method and system, electronic equipment and readable storage medium - Google Patents

Electrocardiosignal classification method and system, electronic equipment and readable storage medium Download PDF

Info

Publication number
CN114190889A
CN114190889A CN202111391932.XA CN202111391932A CN114190889A CN 114190889 A CN114190889 A CN 114190889A CN 202111391932 A CN202111391932 A CN 202111391932A CN 114190889 A CN114190889 A CN 114190889A
Authority
CN
China
Prior art keywords
classification
electrocardiosignals
classification model
electrocardiosignal
signal
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.)
Pending
Application number
CN202111391932.XA
Other languages
Chinese (zh)
Inventor
丑远婷
梁欣然
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.)
Shanghai United Imaging Intelligent Healthcare Co Ltd
Original Assignee
Shanghai United Imaging Intelligent Healthcare Co Ltd
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 Shanghai United Imaging Intelligent Healthcare Co Ltd filed Critical Shanghai United Imaging Intelligent Healthcare Co Ltd
Priority to CN202111391932.XA priority Critical patent/CN114190889A/en
Publication of CN114190889A publication Critical patent/CN114190889A/en
Pending legal-status Critical Current

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]

Landscapes

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

Abstract

The invention discloses a method and a system for classifying electrocardiosignals, electronic equipment and a readable storage medium, wherein the method comprises the following steps: acquiring an electrocardiosignal; inputting the electrocardiosignals into a first classification model, and outputting an initial classification result of the electrocardiosignals; acquiring a characteristic diagram output by a residual error module of the first classification model; and inputting the initial classification result and the characteristic diagram of the electrocardiosignals into a second classification model, and outputting the final classification result of the electrocardiosignals. According to the method and the device, the primary classification result of each electrocardiosignal is obtained by adopting the first classification model, the characteristic diagram extracted by the first classification model is further combined with the primary classification result output by the first classification model for re-analysis, the second classification model is used for classification, the final classification result is obtained, the robustness of the classification model is improved, and the final classification precision of the electrocardiosignals is also improved.

Description

Electrocardiosignal classification method and system, electronic equipment and readable storage medium
Technical Field
The invention belongs to the field of electrocardiosignal identification and classification, and particularly relates to an electrocardiosignal classification method, an electrocardiosignal classification system, an electronic device and a readable storage medium.
Background
Electrocardiographic examination is an important method for clinically diagnosing cardiovascular diseases, and accurate automatic electrocardiographic analysis can provide certain auxiliary information for doctors. With the rapid development of artificial intelligence, deep learning has been successfully applied in various fields by virtue of its powerful feature extraction and classification capabilities, for example, signals are input into a deep learning network for classification, or artificial features are combined as input of the network for classification, but because such methods rely on artificial feature selection, the selected features may have certain limitations and cannot fully express deep characteristics of the signals; some researchers use a deep learning network as a feature extractor to extract high-level abstract features of the electrocardiosignals and then classify the electrocardiosignals, but because the electrocardiosignals are weak signals with non-stationarity and randomness and are easy to interfere, the classification accuracy of the electrocardiosignals still needs to be researched and improved. In summary, there is still much research space for how a deep learning-based electrocardiosignal classification algorithm can achieve higher accuracy.
Disclosure of Invention
The present invention provides a method, a system, an electronic device and a readable storage medium for classifying an electrocardiographic signal, so as to overcome the above-mentioned defects in the prior art.
The invention solves the technical problems through the following technical scheme:
a method of classifying an electrocardiographic signal, comprising:
acquiring an electrocardiosignal;
inputting the electrocardiosignals into a first classification model, and outputting an initial classification result of the electrocardiosignals;
acquiring a characteristic diagram output by a residual error module of the first classification model;
and inputting the initial classification result and the characteristic diagram of the electrocardiosignals into a second classification model, and outputting the final classification result of the electrocardiosignals.
Preferably, the method further comprises the following steps:
and performing dimension reduction processing on the feature map to obtain a feature vector.
Preferably, the obtaining of the feature map output by the residual module of the first classification model includes:
and acquiring a characteristic diagram output by the last residual error module of the first classification model.
Preferably, the method further comprises the following steps:
preprocessing the electrocardiosignals;
the preprocessing comprises extracting the electrocardiosignals to obtain at least one signal segment with preset signal length.
Preferably, the pretreatment specifically comprises:
if the length of the electrocardiosignal is smaller than the preset signal length, 0 is supplemented to the preset signal length to obtain a signal section;
and if the length of the electrocardiosignal is greater than the preset signal length, extracting the electrocardiosignal according to a preset step length to obtain at least one signal section with the preset signal length.
Preferably, the method further comprises the following steps:
obtaining the lead of the electrocardiosignal;
the step of inputting the electrocardiosignals into the first classification model comprises the following steps:
and inputting the electrocardiosignals and the leads to a first classification model.
Preferably, a first classification model in the classification method is obtained by training based on a first training set;
the first training set comprises a plurality of sample electrocardiosignals and a classification label of each sample electrocardiosignal;
and/or the presence of a gas in the gas,
a second classification model in the classification method is obtained by training based on a second training set;
the second training set comprises initial classification results of a plurality of sample electrocardiosignals input into the first classification model, a feature map output by a residual error module of the first classification model after each sample electrocardiosignal is input into the first classification model, and a classification label of each sample electrocardiosignal.
A system for classifying an electrocardiographic signal, comprising:
the signal acquisition module is used for acquiring electrocardiosignals;
the initial classification module is used for inputting the electrocardiosignals into a first classification model and outputting an initial classification result of the electrocardiosignals;
the characteristic diagram obtaining module is used for obtaining a characteristic diagram output by the residual error module of the first classification model;
and the final classification module is used for inputting the initial classification result and the characteristic diagram of the electrocardiosignals into a second classification model and outputting the final classification result of the electrocardiosignals.
An electronic device comprises a memory, a processor and a computer program stored on the memory and capable of running on the processor, wherein the processor executes the computer program to realize the electrocardiosignal classification method.
A computer-readable storage medium, on which a computer program is stored which, when being executed by a processor, carries out the above-mentioned method of classifying an electrocardiographic signal.
The positive progress effects of the invention are as follows: according to the method and the device, the primary classification result of each electrocardiosignal is obtained by adopting the first classification model, the characteristic diagram extracted by the first classification model is further combined with the primary classification result output by the first classification model for re-analysis, the second classification model is used for classification, the final classification result is obtained, the robustness of the classification model is improved, and the final classification precision of the electrocardiosignals is also improved.
Drawings
Fig. 1 is a flowchart of a method for classifying electrocardiographic signals according to embodiment 1 of the present invention.
Fig. 2 is a flowchart of a method for classifying electrocardiographic signals according to embodiment 2 of the present invention.
Fig. 3 is a flowchart of a method for classifying electrocardiographic signals according to embodiment 3 of the present invention.
Fig. 4 is a flowchart of a method for classifying electrocardiographic signals according to embodiment 4 of the present invention.
FIG. 5 is a flowchart of a SE-ResNet based deep convolutional neural network in embodiment 5 of the present invention.
Fig. 6 is a schematic block diagram of a system for classifying electrocardiographic signals according to embodiment 6 of the present invention.
Fig. 7 is a schematic structural diagram of an electronic device according to embodiment 7 of the present invention.
Detailed Description
The invention is further illustrated by the following examples, which are not intended to limit the scope of the invention.
Example 1
The method comprises an electrocardiosignal classification and feature extraction module of a deep convolution neural network based on SE-ResNet, a feature dimension reduction module based on factor analysis and a classification module of a BiLSTM network based on an attention mechanism. The system comprises an electrocardiosignal classification and feature extraction module of a deep convolution neural network based on SE-ResNet, a feature dimension reduction module and a classification module of a BilSTM network based on an attention mechanism, wherein the electrocardiosignal classification and feature extraction module of the deep convolution neural network based on SE-ResNet is used for primarily classifying input electrocardiosignals to obtain an initial classification result and a feature map output by the neural network, the feature dimension reduction module is used for performing dimension reduction processing on the feature map to obtain a feature vector, and the classification module of the BilStTM network based on the attention mechanism is used for analyzing the feature vector and the initial classification result output by a first classification model again to obtain a final signal classification result.
Specifically, as shown in fig. 1, the method for classifying electrocardiographic signals includes:
step 10, acquiring electrocardiosignals;
specifically, the electrocardiograph signals can be obtained through an electrocardiograph, or can be obtained through some intelligent wearable devices such as a bracelet, a watch and the like, the basic principle is that the electrocardiograph signals are obtained through electrodes in contact with the body surface of a human body, along with the medical instrument technology, the number of leads of the electrocardiograph is more and more, from the initial single lead to the three leads, the eighteen leads and the like, twelve leads which are commonly used in the clinical practice of hospitals at present are called as standard leads. The intelligent wearable device can acquire single-lead and multi-lead electrocardiosignals.
Step 20, inputting the electrocardiosignals into a first classification model, and outputting an initial classification result of the electrocardiosignals;
specifically, the first classification model may be a conventional classifier (e.g., SVM, MLP, etc.) or a deep learning network (e.g., RNN, CNN, etc.) to classify the input electrocardiographic signals. In the embodiment, a deep convolutional neural network based on SE-ResNet is preferably constructed to serve as a first classification model, a hidden layer of the deep convolutional neural network is composed of a convolutional layer, a plurality of SE-ResNet modules and a pooling layer, and finally a probability value of the current electrocardiosignal belonging to each class is output through a full connection layer and a softmax function.
Step 30, obtaining a characteristic diagram output by a residual error module of the first classification model;
specifically, the conventional convolutional neural network mainly includes two operations, one is convolution and one is pooling, where a pooling layer does not affect interaction between channels, and only performs pooling operation in each channel, while a convolutional layer can perform information interaction between channels, and then generates a new channel in a next layer, and a feature map is a result output by acting on the convolutional layer, whereas for a residual module, a conventional residual module is generally formed by logically combining a plurality of convolutional layers, and further can output a feature map corresponding to each residual module, and if there are a plurality of residual modules, each residual module has a corresponding feature map output.
And step 40, inputting the initial classification result and the characteristic diagram of the electrocardiosignals into a second classification model, and outputting the final classification result of the electrocardiosignals.
Specifically, the second classification model may be a conventional classifier (e.g., SVM, MLP, etc.) or a deep learning network (e.g., RNN, CNN, etc.), and since the RNN can process inputs of different lengths and can mine the relationship between sequence data, the application preferably uses a BiLSTM network based on an attention mechanism as the second classification model to capture the relationship between electrocardiographic signals, so as to obtain a better final classification result of the electrocardiographic signals.
In this embodiment, the first classification model is used to obtain the preliminary classification result of each electrocardiographic signal, the feature map extracted by the first classification model is further combined with the initial classification result output by the first classification model to perform reanalysis, and the second classification model is used to perform classification to obtain the final classification result.
Example 2
The method for classifying electrocardiographic signals according to this embodiment is further improved on the basis of embodiment 1, as shown in fig. 2, before step 40, the method further includes:
step 31, performing dimension reduction processing on the feature map to obtain a feature vector;
specifically, the feature map obtained through the first classification model has larger dimension and certain redundant information, so that the feature map is preferably subjected to dimension reduction, on one hand, the data dimension can be reduced to accelerate the data processing of the later model, on the other hand, the original features are converted into the features beneficial to classification, and the classification precision is improved. Preferably, the dimension reduction processing is performed on the feature map by adopting a factor analysis method.
Furthermore, the feature vector obtained after the dimension reduction processing is used as a part of the input features of the second classification model to obtain a better final classification result of the electrocardiosignals.
In this embodiment, referring to fig. 2, step 30 specifically includes:
and 301, acquiring a feature map output by the last residual error module of the first classification model.
Specifically, in the residual error modules of the first classification model of the electrocardiographic signals, the output of the last residual error module can further extract high-level abstract features of the electrocardiographic signals, so that a feature map output by the residual error module is preferably taken as a new feature of a subsequently input second classification model, and meanwhile, data dimensionality can be reduced to accelerate the subsequent data processing speed.
In this embodiment, in order to obtain a better new feature, it is preferable to adopt the feature map output by the last residual module for the feature map output by the first classification model, and perform the dimension reduction processing on the feature map, so that the data processing speed and the classification accuracy of the electrocardiographic signal can be comprehensively considered.
Example 3
The method for classifying electrocardiographic signals according to the present embodiment is further improved on the basis of embodiment 1, as shown in fig. 3, after step 10, the method further includes:
step 11, preprocessing the electrocardiosignal;
the preprocessing comprises the steps of extracting the electrocardiosignals to obtain at least one signal section with a preset signal length;
specifically, for electrocardiosignals with different durations, signal segments with the same length are set and extracted, so that the dimensionalities of the signal segments input into the network are consistent, and the data processing of a subsequent model is conveniently completed. For example, 8528 single-lead electrocardiosignals are obtained in total, the signal sampling rate is 300Hz, the time duration of each electrocardiosignal is different, and the distribution range is approximately 9s to 61 s. For each electrocardiosignal, a signal segment of 10s (preset signal length) is sequentially extracted, and the signal length per second corresponds to 300 sampling points.
Further, based on the obtained signal segment, the signal segment is used as input data and is input into a first classification model; and inputting the initial classification result and the characteristic diagram of each signal segment output by the first classification model into the second classification model, and further outputting the final classification result of the electrocardiosignals.
Wherein the pretreatment specifically comprises:
if the length of the electrocardiosignal is smaller than the preset signal length, 0 is supplemented to the preset signal length to obtain a signal section;
specifically, in order to ensure that the dimensions of signals input into the network are consistent, for an electrocardiograph signal with a short signal length, for example, if the length is less than 10s, 0 is supplemented to a sampling point to a preset signal length, so as to complete data processing of a subsequent model.
And if the length of the electrocardiosignal is greater than the preset signal length, extracting the electrocardiosignal according to a preset step length to obtain at least one signal section with the preset signal length.
Specifically, if the length exceeds 10S, for each electrocardiograph signal, a signal segment of 10S (preset signal length) is extracted every 20S (preset step length), for example, if the length is less than 20S, a 10S signal segment is obtained, and if the length is 29S, after 1 10S signal segment is extracted, only 9S data is obtained in the next step length, 0 to 10S are supplemented to the data, and a 2 nd signal segment is obtained.
In this embodiment, in order to realize classification of electrocardiographic signals with different durations and improve the classification accuracy, original electrocardiographic signals are segmented, so that the dimensions of signal segments input to a network are consistent, and data processing of a subsequent model is completed.
Example 4
The method for classifying electrocardiographic signals according to this embodiment is further improved on the basis of embodiment 1, as shown in fig. 4, before step 20, the method further includes:
step 12, obtaining the lead of the electrocardiosignal;
specifically, the actually acquired data may be data of a plurality of leads, and for electrocardiographic signals of different leads, because the data correspond to different classification label ranges, for example, classification of normal, atrial fibrillation, other rhythms and noise is performed on a lead I signal, and classification of normal, abnormal and noise is performed on a lead II signal, it is necessary to determine which lead corresponds to the lead during classification, so that the lead of the electrocardiographic signal needs to be acquired before classification.
Further, step 20 comprises:
step 202, inputting the electrocardiosignals and the leads to which the electrocardiosignals belong into a first classification model, and outputting an initial classification result of the electrocardiosignals.
Specifically, the lead to which the user belongs is input into the first classification model, and a selection module may be arranged in the first classification model in advance, and is used to select the corresponding classification label range and the corresponding classification model, or when parameter setting is performed in the first classification model, the lead to which the user belongs is used as a parameter of the first classification model, so as to classify the electrocardiographic signals of different leads.
In this embodiment, it is considered that actually acquired data of a plurality of leads and electrocardiographic signals of different leads should have different classification label ranges, so that the leads of the electrocardiographic signals need to be determined during primary classification, and further refinement and improvement of classification results of the electrocardiographic signals are required.
Example 5
In this embodiment, a first classification model is obtained by training based on a first training set, a second classification model is obtained by training based on a second training set, and the classification model is further applied to the classification method for electrocardiographic signals according to embodiment 1;
the first training set comprises a plurality of sample electrocardiosignals and a classification label of each sample electrocardiosignal;
the second training set comprises initial classification results of a plurality of sample electrocardiosignals input into the first classification model, a feature map output by a residual error module of the first classification model after each sample electrocardiosignal is input into the first classification model, and a classification label of each sample electrocardiosignal.
Specifically, in the training, the sample electrocardiographic signals or the preprocessed signals (such as signal denoising and data enhancement) may be directly input into a deep learning network (such as RNN, CNN, etc.) to be trained to obtain a first classification model and a second classification model, or the deep learning network may be used as a feature extractor to extract high-level abstract features of the sample electrocardiographic signals, and then the training is performed by using a conventional classification model to obtain the first classification model and the second classification model.
Or, the sample electrocardiographic signal is divided into a plurality of signal segments with equal size, and then training is performed based on each signal segment, in addition, the lead to which the sample electrocardiographic signal belongs can also be used as a parameter of the model, so as to further realize classification of the electrocardiographic signals of different leads, specifically, the training sets of the first classification model and the second classification model are as follows:
the first training set comprises a plurality of sample signal segments of a plurality of sample electrocardiosignals, leads of each sample electrocardiosignal and a classification label of each sample signal segment;
the sample signal segment is obtained by extracting the sample electrocardiosignals according to a preset step length;
in the training process of the first classification model, in order to enable more data to participate in training and balance different classes of data, the step size is set to different values to increase the data. Taking lead I data as an example, the step size is set to 1000 samples when the signal category is noise and atrial fibrillation, and 1500 samples when the signal category is normal and other rhythms.
The second training set comprises initial classification results after a plurality of sample signal segments are input into the first classification model, feature maps output by a residual error module of the first classification model after each sample signal segment is input into the first classification model, and classification labels of each signal segment;
and the sample signal section is obtained by extracting the sample electrocardiosignals according to a preset step length.
In the training and verification process of the second classification model, in order to ensure the integrity of data, the step length can be set to 1000 sampling points, and no data is discarded. Based on the setting, inputting the obtained multiple sample signal segments into the first classification model to obtain corresponding initial classification results, and then training a second classification model based on the initial classification results and the feature diagram output by the residual error module as new features.
Specifically, the data training process is as follows, taking a single lead electrocardiosignal as an example:
1. data pre-processing
Acquiring a lead I electrocardiosignal x, segmenting an original signal, and if the length of the original signal is less than 10s, performing 0 supplementation operation on the original signal; if the length of the original signal is greater than 10s, the original signal is segmented and extracted according to the step length, the length of each segment of the signal is 10s, and the step length is 1/3 of the length of the signal segment (the step length can be dynamically adjusted according to the actual length of the signal). Thus, a segment of the original signal x is extracted to obtain l signal segments, x ═ x1,x2,…,xi,…,xlEach signal segment xiLabel y ofiAre consistent with the label y of the original signal, and the label is the classification result of the known electrocardiosignals. Wherein, the classification label range of the lead I is as follows: normal (Normal), Atrial Fibrillation (AF), Other rhythms (Other) and noise signals (noise).
In addition, when the acquired training set data is processed, if the number of normal electrocardiosignals is large, for example, more than 90%, the data can be filtered to balance the data, and meanwhile, the problem that the quality of the initial section of signals is poor due to instability of the signals when the data are just acquired can be avoided, the previous sampling data can be discarded, and specifically, for the electrocardiosignals of which the length of the electrocardiosignals is greater than a certain length threshold, the previous section of signals in the electrocardiosignals is filtered and used for training the first model. For example, if the length of the electrocardiographic signal data exceeds 9000 sampling points (the length of the 9000 sampling points is 30s, and 30s is set as a length threshold), the last 8000 data points of the whole signal should contain most information of the signal, and the first 1000 data points are filtered.
2. Feature extraction and classification
Constructing a deep convolution neural network based on SE-ResNet as a classifier, wherein the input of the network is a signal segment x of 10s as shown in figure 5iThe hidden layer in the middle is composed of a convolutional layer (convergence), a normalization layer (Bath Norm), an activation function layer (Relu), a max pooling layer (MaxPooling), a plurality of residual modules (SE-ResNet Module) and a global pooling layer (GlobalPooling), wherein x 3' in the figure represents that three groups of structures are composed of 2 residual modules and 1 max pooling layer and are sequentially executed, and finally, the probability value of each category of the current signal segment is output through a full connection layer (FC) and a softmax function. Label y of each signal segmentiAs a supervisory signal, an optimization training is performed using a training set. Obtaining a first classification model after the model learning is completed, and using the first classification model to perform classification on each signal segment xiAre classified to obtain
Figure BDA0003364146650000101
And the feature map F output by the last SE-ResNet modulei∈R8×84。
3. Feature dimension reduction and combination
Using factor analysis to profile FiDimension reduction is carried out, in this example, the number of the sampling factors is set to be 20 to obtain F'i∈R1 ×20On the one hand, the data dimension can be reduced to accelerate the later modelOn the other hand, the original features can be converted into the features beneficial to classification, so that the classification precision is improved. In order to improve the classification performance of the subsequent network and reduce the information loss, the classification results of each signal segment are combined
Figure BDA0003364146650000102
And post-dimensionality reduction characteristic F'iGenerating a new signature M ═ M1,M2,…,Ml]T. For example
Figure BDA0003364146650000103
F′i＝[2,2,…,2,2]Then taken together are Mi＝[2,2,…,2,1,0,0,0]。
4. Signal classification
Because the RNN can process inputs of different lengths, and can mine the relationship between sequence data. Therefore, a BilSt network based on an attention mechanism is used as a classifier to capture the relation among signal segments, and the input of the network is M ═ M obtained by the electrocardiosignal after the segmentation, classification and feature combination1,M2,…,Ml]TAnd l is the number of the signal segments. The supervision signal of the network is the whole electrocardiosignal label y. And (4) obtaining a second classification model after finishing supervised learning on the training set, wherein the second classification model can realize the electrocardiosignal classification function. In this example, the model is further verified based on five cross-verifications, resulting in an average result F1 score up to 0.8477.
Example 6
A system for classifying electrocardiographic signals, as shown in fig. 6, includes:
the signal acquisition module 1 is used for acquiring electrocardiosignals;
specifically, the electrocardiograph signals can be obtained through an electrocardiograph, or can be obtained through some intelligent wearable devices such as a bracelet, a watch and the like, the basic principle is that the electrocardiograph signals are obtained through electrodes in contact with the body surface of a human body, along with the medical instrument technology, the number of leads of the electrocardiograph is more and more, from the initial single lead to the three leads, the eighteen leads and the like, twelve leads which are commonly used in the clinical practice of hospitals at present are called as standard leads. The intelligent wearable device can acquire single-lead and multi-lead electrocardiosignals.
The initial classification module 2 is used for inputting the electrocardiosignals into a first classification model and outputting initial classification results of the electrocardiosignals;
specifically, the first classification model may be a conventional classifier (e.g., SVM, MLP, etc.) or a deep learning network (e.g., RNN, CNN, etc.) to classify the input electrocardiographic signals. In the embodiment, a deep convolutional neural network based on SE-ResNet is preferably constructed to serve as a first classification model, a hidden layer of the deep convolutional neural network is composed of a convolutional layer, a plurality of SE-ResNet modules and a pooling layer, and finally a probability value of the current electrocardiosignal belonging to each class is output through a full connection layer and a softmax function.
A feature map obtaining module 3, configured to obtain a feature map output by a residual module of the first classification model;
specifically, the conventional convolutional neural network mainly includes two operations, one is convolution and one is pooling, where a pooling layer does not affect interaction between channels, and only performs pooling operation in each channel, while a convolutional layer can perform information interaction between channels, and then generates a new channel in a next layer, and a feature map is a result output by acting on the convolutional layer, whereas for a residual module, a conventional residual module is generally formed by logically combining a plurality of convolutional layers, and further can output a feature map corresponding to each residual module, and if there are a plurality of residual modules, each residual module has a corresponding feature map output.
And the final classification module 4 is used for inputting the initial classification result and the characteristic diagram of the electrocardiosignals into a second classification model and outputting the final classification result of the electrocardiosignals.
Specifically, the second classification model may be a conventional classifier (e.g., SVM, MLP, etc.) or a deep learning network (e.g., RNN, CNN, etc.), and since the RNN can process inputs of different lengths and can mine the relationship between sequence data, the application preferably uses a BiLSTM network based on an attention mechanism as the second classification model to capture the relationship between electrocardiographic signals, so as to obtain a better final classification result of the electrocardiographic signals.
In this embodiment, the first classification model is used to obtain the preliminary classification result of each electrocardiographic signal, the feature map extracted by the first classification model is further combined with the initial classification result output by the first classification model to perform reanalysis, and the second classification model is used to perform classification to obtain the final classification result.
Example 7
An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the method for classifying an electrocardiographic signal according to any one of embodiments 1 to 5 when executing the computer program.
Fig. 7 is a schematic structural diagram of an electronic device provided in this embodiment. FIG. 7 illustrates a block diagram of an exemplary electronic device 90 suitable for use in implementing embodiments of the present invention. The electronic device 90 shown in fig. 7 is only an example, and should not bring any limitation to the functions and the scope of use of the embodiment of the present invention.
As shown in fig. 7, the electronic device 90 may be embodied in the form of a general purpose computing device, which may be, for example, a server device. The components of the electronic device 90 may include, but are not limited to: at least one processor 91, at least one memory 92, and a bus 93 that connects the various system components (including the memory 92 and the processor 91).
The bus 93 includes a data bus, an address bus, and a control bus.
Memory 92 may include volatile memory, such as Random Access Memory (RAM)921 and/or cache memory 922, and may further include Read Only Memory (ROM) 923.
Memory 92 may also include a program tool 925 having a set (at least one) of program modules 924, such program modules 924 including, but not limited to: an operating system, one or more application programs, other program modules, and program data, each of which, or some combination thereof, may comprise an implementation of a network environment.
The processor 91 executes various functional applications and data processing by running a computer program stored in the memory 92.
The electronic device 90 may also communicate with one or more external devices 94 (e.g., keyboard, pointing device, etc.). Such communication may be through an input/output (I/O) interface 95. Also, the electronic device 90 may communicate with one or more networks (e.g., a Local Area Network (LAN), a Wide Area Network (WAN), and/or a public network, such as the Internet) via a network adapter 96. The network adapter 96 communicates with the other modules of the electronic device 90 via the bus 93. It should be appreciated that although not shown in the figures, other hardware and/or software modules may be used in conjunction with the electronic device 90, including but not limited to: microcode, device drivers, redundant processors, external disk drive arrays, RAID (disk array) systems, tape drives, and data backup storage systems, etc.
It should be noted that although in the above detailed description several units/modules or sub-units/modules of the electronic device are mentioned, such a division is merely exemplary and not mandatory. Indeed, the features and functionality of two or more of the units/modules described above may be embodied in one unit/module, according to embodiments of the application. Conversely, the features and functions of one unit/module described above may be further divided into embodiments by a plurality of units/modules.
Example 7
A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, implements the method for classifying an electrocardiographic signal according to any one of embodiments 1 to 5.
More specific examples, among others, that the readable storage medium may employ may include, but are not limited to: a portable disk, a hard disk, random access memory, read only memory, erasable programmable read only memory, optical storage device, magnetic storage device, or any suitable combination of the foregoing.
In a possible implementation manner, the present invention can also be implemented in the form of a program product, which includes program code for causing a terminal device to execute a method for classifying electrocardiographic signals according to any one of embodiments 1 to 5 when the program product runs on the terminal device.
Where program code for carrying out the invention is written in any combination of one or more programming languages, the program code may be executed entirely on the user device, partly on the user device, as a stand-alone software package, partly on the user device and partly on a remote device or entirely on the remote device.
While specific embodiments of the invention have been described above, it will be appreciated by those skilled in the art that this is by way of example only, and that the scope of the invention is defined by the appended claims. Various changes and modifications to these embodiments may be made by those skilled in the art without departing from the spirit and scope of the invention, and these changes and modifications are within the scope of the invention.

Claims (10)

1. A method for classifying an electrocardiosignal, comprising:
acquiring an electrocardiosignal;
inputting the electrocardiosignals into a first classification model, and outputting an initial classification result of the electrocardiosignals;
acquiring a characteristic diagram output by a residual error module of the first classification model;
and inputting the initial classification result and the characteristic diagram of the electrocardiosignals into a second classification model, and outputting the final classification result of the electrocardiosignals.
2. The method for classifying electrocardiographic signals according to claim 1, further comprising:
and performing dimension reduction processing on the feature map to obtain a feature vector.
3. The method for classifying electrocardiographic signals according to claim 1, wherein the obtaining of the feature map output by the residual module of the first classification model comprises:
and acquiring a characteristic diagram output by the last residual error module of the first classification model.
4. The method for classifying electrocardiographic signals according to claim 1, further comprising:
preprocessing the electrocardiosignals;
the preprocessing comprises extracting the electrocardiosignals to obtain at least one signal segment with preset signal length.
5. The method for classifying electrocardiographic signals according to claim 4, wherein the preprocessing specifically comprises:
if the length of the electrocardiosignal is smaller than the preset signal length, 0 is supplemented to the preset signal length to obtain a signal section;
and if the length of the electrocardiosignal is greater than the preset signal length, extracting the electrocardiosignal according to a preset step length to obtain at least one signal section with the preset signal length.
6. The method for classifying electrocardiographic signals according to claim 1, further comprising:
obtaining the lead of the electrocardiosignal;
the step of inputting the electrocardiosignals into the first classification model comprises the following steps:
and inputting the electrocardiosignals and the leads to a first classification model.
7. The method for classifying cardiac signals according to claim 1, wherein a first classification model in the classification method is trained based on a first training set;
the first training set comprises a plurality of sample electrocardiosignals and a classification label of each sample electrocardiosignal;
and/or the presence of a gas in the gas,
a second classification model in the classification method is obtained by training based on a second training set;
the second training set comprises initial classification results of a plurality of sample electrocardiosignals input into the first classification model, a feature map output by a residual error module of the first classification model after each sample electrocardiosignal is input into the first classification model, and a classification label of each sample electrocardiosignal.
8. A system for classifying an electrocardiographic signal, comprising:
the signal acquisition module is used for acquiring electrocardiosignals;
the initial classification module is used for inputting the electrocardiosignals into a first classification model and outputting an initial classification result of the electrocardiosignals;
the characteristic diagram obtaining module is used for obtaining a characteristic diagram output by the residual error module of the first classification model;
and the final classification module is used for inputting the initial classification result and the characteristic diagram of the electrocardiosignals into a second classification model and outputting the final classification result of the electrocardiosignals.
9. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the method for classifying an electrocardiographic signal according to any one of claims 1 to 7 when executing the computer program.
10. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the method for classifying an electrocardiographic signal according to any one of claims 1 to 7.
CN202111391932.XA 2021-11-19 2021-11-19 Electrocardiosignal classification method and system, electronic equipment and readable storage medium Pending CN114190889A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202111391932.XA CN114190889A (en) 2021-11-19 2021-11-19 Electrocardiosignal classification method and system, electronic equipment and readable storage medium

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202111391932.XA CN114190889A (en) 2021-11-19 2021-11-19 Electrocardiosignal classification method and system, electronic equipment and readable storage medium

Publications (1)

Publication Number Publication Date
CN114190889A true CN114190889A (en) 2022-03-18

Family

ID=80648401

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202111391932.XA Pending CN114190889A (en) 2021-11-19 2021-11-19 Electrocardiosignal classification method and system, electronic equipment and readable storage medium

Country Status (1)

Country Link
CN (1) CN114190889A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117763399A (en) * 2024-02-21 2024-03-26 电子科技大学 Neural network classification method for self-adaptive variable-length signal input

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117763399A (en) * 2024-02-21 2024-03-26 电子科技大学 Neural network classification method for self-adaptive variable-length signal input
CN117763399B (en) * 2024-02-21 2024-05-14 电子科技大学 Neural network classification method for self-adaptive variable-length signal input

Similar Documents

Publication Publication Date Title
WO2021051875A1 (en) Cell classification method and apparatus, medium and electronic device
CN111275093A (en) Heart beat classification method and device for multi-label labeling electrocardiosignals
Huang et al. End-to-end continuous emotion recognition from video using 3D ConvLSTM networks
CN110796199A (en) Image processing method and device and electronic medical equipment
CN111368878B (en) Optimization method based on SSD target detection, computer equipment and medium
CN112270240B (en) Signal processing method, device, electronic equipment and storage medium
CN110400610B (en) Small sample clinical data classification method and system based on multichannel random forest
CN112365980A (en) Brain tumor multi-target point auxiliary diagnosis and prospective treatment evolution visualization method and system
CN115363599B (en) Electrocardiosignal processing method and system for atrial fibrillation recognition
Ullah et al. An End‐to‐End Cardiac Arrhythmia Recognition Method with an Effective DenseNet Model on Imbalanced Datasets Using ECG Signal
CN111956214A (en) QRS wave automatic detection method based on U-net end-to-end neural network model
US20230101539A1 (en) Physiological electric signal classification processing method and apparatus, computer device and storage medium
CN111358460A (en) Arrhythmia identification method and device and electronic equipment
CN114190889A (en) Electrocardiosignal classification method and system, electronic equipment and readable storage medium
US11410016B2 (en) Selective performance of deterministic computations for neural networks
CN110781849A (en) Image processing method, device, equipment and storage medium
CN112837466B (en) Bill recognition method, device, equipment and storage medium
Lu et al. Image recognition algorithm based on improved AlexNet and shared parameter transfer learning
CN112244863A (en) Signal identification method, signal identification device, electronic device and readable storage medium
CN117173697A (en) Cell mass classification and identification method, device, electronic equipment and storage medium
CN114783597B (en) Method and device for diagnosing multi-class diseases, electronic equipment and storage medium
CN113591791B (en) Lung cancer automatic identification system based on self-learning artificial intelligence
Liu et al. Adaptive shapelets preservation for time series augmentation
Ma Summary of research on application of deep learning in image recognition
CN113723519A (en) Electrocardio data processing method and device based on contrast learning and storage medium

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