CN110037682B - Method for recognizing heart rhythm type based on improved convolutional neural network - Google Patents
Method for recognizing heart rhythm type based on improved convolutional neural network Download PDFInfo
- Publication number
- CN110037682B CN110037682B CN201910257087.3A CN201910257087A CN110037682B CN 110037682 B CN110037682 B CN 110037682B CN 201910257087 A CN201910257087 A CN 201910257087A CN 110037682 B CN110037682 B CN 110037682B
- Authority
- CN
- China
- Prior art keywords
- neural network
- convolutional
- atrial fibrillation
- electrocardiosignals
- layer
- 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.)
- Active
Links
Images
Classifications
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
- A61B5/316—Modalities, i.e. specific diagnostic methods
- A61B5/318—Heart-related electrical modalities, e.g. electrocardiography [ECG]
-
- 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
- A61B5/361—Detecting fibrillation
-
- 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
Landscapes
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Engineering & Computer Science (AREA)
- Cardiology (AREA)
- Physics & Mathematics (AREA)
- Biophysics (AREA)
- Surgery (AREA)
- Artificial Intelligence (AREA)
- Veterinary Medicine (AREA)
- Public Health (AREA)
- General Health & Medical Sciences (AREA)
- Animal Behavior & Ethology (AREA)
- Molecular Biology (AREA)
- Pathology (AREA)
- Biomedical Technology (AREA)
- Heart & Thoracic Surgery (AREA)
- Medical Informatics (AREA)
- Evolutionary Computation (AREA)
- Mathematical Physics (AREA)
- Fuzzy Systems (AREA)
- Signal Processing (AREA)
- Psychiatry (AREA)
- Physiology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)
Abstract
The present application relates to a method, computer device and storage medium for recognizing a heart rhythm type based on an improved convolutional neural network. Method of producing a composite materialThe method comprises the steps of adopting a pre-trained neural network to identify an electrocardiosignal so as to identify a heart rhythm type corresponding to the electrocardiosignal and determine whether the electrocardiosignal is atrial fibrillation. The pre-trained neural network adopts a loss function as follows:
the loss function is used for increasing the punishment of the convolutional neural network on false negative in training, so that the missing rate is reduced and the accuracy is improved on the premise of ensuring the accuracy.
Description
Technical Field
The application belongs to the technical field of electrocardiogram processing, and particularly relates to a method for recognizing a heart rhythm type based on an improved convolutional neural network.
Background
An Electrocardiogram (ECG) is a graph formed from the surface recording of the changes in electrical activity produced by the heart each cardiac cycle. A plurality of heart diseases of people can be characterized through electrocardiograms. Atrial fibrillation (abbreviated as atrial fibrillation) is the most common persistent arrhythmia. The current classification of atrial fibrillation is not unified, and the identification of atrial fibrillation is complex.
Convolutional Neural Networks (CNN) are a class of feed forward Neural Networks (fed forward Networks) that contain convolution computations and have a deep structure. In recent years, neural networks such as convolutional neural networks are increasingly used to identify types of atrial fibrillation in electrocardiograms. For example, in the dawn gay master thesis "study of atrial fibrillation recognition algorithm based on machine learning", a method for performing atrial fibrillation recognition using a convolutional neural network is explicitly described. In the korean master paper "atrial fibrillation detection based on atrial activity characteristics and convolutional neural network", a method for identifying atrial fibrillation by using a convolutional neural network is also studied, however, in all of the papers, the existing convolutional neural network is directly applied to atrial fibrillation identification, and no optimization is performed on the identification of atrial fibrillation, which results in that the accuracy of atrial fibrillation identification in a electrocardiogram cannot be further improved.
Disclosure of Invention
The technical problem to be solved by the invention is as follows: the method for recognizing the heart rhythm type based on the improved convolutional neural network is high in accuracy rate and low in omission rate of atrial fibrillation recognition in the electrocardiogram.
The technical scheme adopted by the invention for solving the technical problems is as follows:
a method for identifying a heart rhythm type based on an improved convolutional neural network, comprising the steps of:
acquiring an electrocardiosignal to be identified;
preprocessing the electrocardiosignals to obtain electrocardiosignals meeting the identification requirements;
inputting the preprocessed electrocardiosignals into a pre-trained neural network for recognition, and outputting the heart rhythm type corresponding to the electrocardiosignals, wherein the neural network adopts a loss function as follows:
wherein c represents the loss size, n represents the number of samples, x is the sample, y is the label corresponding to the sample x, a is the output size calculated by taking the sample x as the input and using the current neural network, and α are all regularization coefficients.
Preferably, in the method for identifying the heart rhythm type based on the improved convolutional neural network, α is 0.8-0.9, and β is 1.1-1.2 in the loss function.
Preferably, the method of the present invention for identifying a heart rhythm type is based on an improved convolutional neural network comprising 1 input layer, a plurality of convolutional and pooling layers, 1 fully-connected layer and 1 classifier layer.
Preferably, in the method for recognizing a heart rhythm type based on the improved convolutional neural network, when the electrocardiosignal is preprocessed, a filter with a preset cut-off frequency is adopted to filter the electrocardiosignal.
Preferably, the method for identifying the heart rhythm type based on the improved convolutional neural network of the present invention preprocesses the cardiac signal, including:
judging whether the sampling frequency of the electrocardiosignals is a preset frequency or not;
and when the sampling frequency is not the preset frequency, resampling the electrocardiosignals into the electrocardiosignals with the preset frequency by adopting an interpolation method.
Preferably, the method for recognizing a heart rhythm type based on the improved convolutional neural network of the present invention, wherein the pre-trained neural network is obtained by the following steps:
s31: acquiring a training database, wherein the training database is electrocardio data known as atrial fibrillation or non-atrial fibrillation;
s32: preprocessing the electrocardiogram data;
s33: the method is characterized in that a convolutional neural network is used for training, the convolutional neural network is composed of a plurality of convolutional layers, a plurality of pooling layers, a full-link layer and a classifier layer, and known atrial fibrillation or non-atrial fibrillation results are input into the classifier layer during training.
Preferably, the method for identifying the type of heart rhythm based on the improved convolutional neural network of the present invention,
in the step of S31, at least 1 ten thousand of atrial fibrillation electrocardiosignals with 10S and at least 1 ten thousand of other types of electrocardiosignals which are uniformly mixed are used as training data to form a training database, and 0 and 1 are respectively used as labels of the atrial fibrillation electrocardiosignals and the non-atrial fibrillation electrocardiosignals;
in the step of S32, filtering the electrocardio data by using fir filters with upper and lower cut-off frequencies of 0.1Hz and 100Hz respectively, and if the sampling frequency of the electrocardio signals is not 500Hz, resampling the electrocardio signals to 500Hz by using a nearest neighbor interpolation method;
in the step S33, in an 8-layer network, the convolutional layer in layer1 includes 5 kernels, the sizes of the convolutional kernels are all 224, and the step size and the kernel size in the pooling layer in layer1 are both 2; the layer2 convolutional layer contains 5 kernels, the sizes of the convolutional kernels are all 112, and the step size and the kernel size in the pooling layer in the layer2 are both 2; the layer3 convolutional layer comprises 10 kernels, the sizes of the convolutional kernels are both 100, and the step size and the kernel size in the pooling layer in the layer3 are both 2; the layer4 convolutional layer contains 10 kernels, the sizes of the convolutional kernels are both 50, and the step size and the kernel size in the pooling layer in the layer4 are both 2; the layer5 convolutional layer contains 10 kernels, the sizes of the convolutional kernels are both 48, and the step size and the kernel size in the pooling layer in the layer5 are both 2; the layer6 convolutional layer contains 10 kernels, the sizes of the convolutional kernels are both 24, and the step size and the kernel size in the pooling layer in the layer4 are both 2; the number of output features of layer6 is 30, and finally 10 features are output after calculation of a fully connected layer 7.
The invention also provides a method for identifying the type of the heart rhythm based on the improved convolutional neural network, which comprises the following steps:
acquiring an electrocardiosignal to be identified;
preprocessing the electrocardiosignals to obtain electrocardiosignals meeting the identification requirements;
inputting the preprocessed electrocardiosignals into a first neural network trained in advance for recognition;
the pre-trained first neural network uses a loss function as:
wherein c represents the loss size, n represents the number of samples, x represents the samples, y represents the label corresponding to the samples x, a represents the output size calculated by taking the samples x as input and using the current neural network, and α are all regularization coefficients, wherein α is 1-1.2, β is 0.8-1 in the loss function, and the heart rhythm type corresponding to the electrocardiosignal identified by the first neural network is output;
if the heart rhythm type is non-atrial fibrillation, the electrocardiosignal is considered to be non-atrial fibrillation;
if the heart rhythm type is atrial fibrillation, inputting the preprocessed electrocardiosignals into a pre-trained second neural network again for recognition;
the pre-trained second neural network uses a loss function as:
wherein c represents the loss, n represents the number of samples, x represents the samples, y represents the label corresponding to the samples x, a represents the output calculated by using the samples x as input and using the current neural network, and α are all regularization coefficients, α is 0.8-1, β is 1.0-1.2 in the loss function, the heart rhythm type corresponding to the electrocardiosignals identified by the second neural network is output, and the heart rhythm type identified by the second neural network is the heart rhythm type of the electrocardiosignals
Preferably, the method for identifying the heart rhythm type based on the improved convolutional neural network of the present invention, the first neural network uses α ═ 1.2 and β ═ 1 in the loss function.
Preferably, the method for identifying the heart rhythm type based on the improved convolutional neural network of the present invention, the second neural network uses α ═ 0.8 and β ═ 1.2 in the loss function.
The invention has the beneficial effects that:
in the embodiment of the invention, a pre-trained neural network is adopted to identify the electrocardiosignals so as to identify the heart rhythm type corresponding to the electrocardiosignals and determine whether the electrocardiosignals are atrial fibrillation. Specifically, by taking the loss function as:
the loss function is used for increasing the punishment of the convolutional neural network on false negative in training, so that the missing rate is reduced and the accuracy is improved on the premise of ensuring the accuracy.
Drawings
The technical solution of the present application is further explained below with reference to the drawings and the embodiments.
Fig. 1 is a flowchart of a method for identifying a heart rhythm type based on an improved convolutional neural network according to an embodiment of the present application;
FIG. 2 is a schematic diagram of an improved convolutional neural network of an embodiment of the present application;
Detailed Description
It should be noted that the embodiments and features of the embodiments in the present application may be combined with each other without conflict.
The technical solutions of the present application will be described in detail below with reference to the accompanying drawings in conjunction with embodiments.
Examples
The embodiment provides a method for identifying a heart rhythm type based on an improved convolutional neural network, as shown in fig. 1, comprising the following steps:
s1: acquiring an electrocardiosignal to be identified;
s2: preprocessing the electrocardiosignals to obtain electrocardiosignals meeting the identification requirements;
s3: inputting the preprocessed electrocardiosignals into a pre-trained neural network for recognition, and outputting the heart rhythm type corresponding to the electrocardiosignals, wherein the neural network adopts a loss function as follows:
wherein c denotes a loss size, n denotes a number of samples, x denotes a sample, y denotes a label corresponding to the sample x, a denotes an output size calculated by using the sample x as an input and using the current neural network, and α are all regularization coefficients, and α is 1.1 to 1.2, and β is 0.8 to 0.9 in the loss function.
It should be noted that the length of the electrocardiographic signal should be the same as the electrocardiographic signal in the training database when the neural network is trained in advance, and if the length of the electrocardiographic signal is different from the length of the electrocardiographic signal in the training database when the neural network is trained in advance, the electrocardiographic signal is cut to be the same as the electrocardiographic signal in the training database when the neural network is trained in advance.
As an alternative embodiment, the electrocardiographic data is preprocessed in step S2; and filtering the electrocardiosignal by adopting a filter with a preset cut-off frequency.
The method can also be used for preprocessing the sampling frequency of the electrocardiosignals, and comprises the following steps:
judging whether the sampling frequency of the electrocardiosignals is a preset frequency or not;
and when the sampling frequency is not the preset frequency, resampling the electrocardiosignals into the electrocardiosignals with the preset frequency by adopting an interpolation method.
The method also comprises the step of carrying out segmentation operation on the electrocardiosignals to be identified, such as dividing the electrocardiosignals into 10s of atrial fibrillation electrocardiosignals.
Such as: during pretreatment, filtering is carried out by using fir filters with upper and lower cut-off frequencies of 0.1Hz and 100Hz respectively, and if the sampling frequency of the electrocardiosignals is not 500Hz, the electrocardiosignals are resampled to be 500Hz by using a nearest neighbor interpolation method.
The method for recognizing the heart rhythm type based on the improved convolutional neural network of the embodiment needs a pre-trained neural network before using the method for recognizing the heart rhythm type based on the improved convolutional neural network, and the pre-trained method comprises the following steps:
s31: acquiring a training database, wherein the training database is electrocardio data known as atrial fibrillation or non-atrial fibrillation, the length of the electrocardio data can be 10s, preferably 4s, and the time of 4s at least can comprise 6 heartbeats, so that the recognition efficiency can be improved under the condition of ensuring the accuracy, and the time for intercepting the electrocardiosignals to be recognized is equal to the length of the electrocardio data when training a neural network;
at least 1 ten thousand 10s atrial fibrillation electrocardiosignals and at least 1 ten thousand other types of electrocardiosignals which are uniformly mixed are used as training data to form a training database, wherein 0 and 1 are respectively used as labels of the atrial fibrillation electrocardiosignals and the non-atrial fibrillation electrocardiosignals;
s32: preprocessing the electrocardiogram data;
during pretreatment, filtering is carried out by using fir filters with upper and lower cut-off frequencies of 0.1Hz and 100Hz respectively, and if the sampling frequency of the electrocardiosignals is not 500Hz, the electrocardiosignals are resampled to be 500Hz by using a nearest neighbor interpolation method.
S33: with 8 layersTraining a Convolutional Neural Network (CNN) of the network, wherein the Convolutional Neural Network (CNN) has a structure that the 1 st layer to the 6 th layer (layer1-layer6) in an 8-layer network are all composed of a convolutional layer and a pooling layer; the convolutional layer in layer1 contains 5 kernels, the sizes of the convolutional kernels are all 224, and the step size and the kernel size in the pooling layer in layer1 are both 2; the layer2 convolutional layer contains 5 kernels, the sizes of the convolutional kernels are all 112, and the step size and the kernel size in the pooling layer in the layer2 are both 2; the layer3 convolutional layer comprises 10 kernels, the sizes of the convolutional kernels are both 100, and the step size and the kernel size in the pooling layer in the layer3 are both 2; the layer4 convolutional layer contains 10 kernels, the sizes of the convolutional kernels are both 50, and the step size and the kernel size in the pooling layer in the layer4 are both 2; the layer5 convolutional layer contains 10 kernels, the sizes of the convolutional kernels are both 48, and the step size and the kernel size in the pooling layer in the layer5 are both 2; the layer6 convolutional layer contains 10 kernels, the sizes of the convolutional kernels are both 24, and the step size and the kernel size in the pooling layer in the layer4 are both 2; the number of input neurons of the fully-connected layer (layer7) is consistent with the number of output features of layer6, and the following formula is used as a loss function in training:
the number of output features of layer6 is 30, and finally 10 features are output after full-connection layer calculation. The 10 features output by the full link layer are input into a classifier layer (layer8) as input, and the known results of atrial fibrillation or non-atrial fibrillation are input for training, wherein the training algorithm can adopt any existing training algorithm. The training algorithm may be: a random gradient descent algorithm, an Adam algorithm, a RMSProp algorithm, an adagard algorithm, an adapelta algorithm, an Adamax algorithm, and the like.
In the loss function, c represents the loss magnitude, n represents the number of samples, x is the sample, y is the label corresponding to sample x, and a is the output magnitude calculated by taking sample x as input and using the current neural network, wherein α are all regularization coefficients
α is 0.8-0.9, β is 1.1-1.2. the loss function is a function for calculating loss, the loss is the difference between the predicted value and the true value of a single sample, and the loss function can be used for increasing the punishment of the convolutional neural network on false negative during training, so that the missing rate is reduced and the accuracy is improved on the premise of ensuring the accuracy.
When the heart rhythm type is identified, the preprocessed electrocardiosignals to be identified are input into a neural network, 10 features are output after being processed by layer1-7 of the neural network trained in advance, the 10 features are input into a classifier layer (layer8), and layer8 calculates the output result of the electrocardiosignals to be identified, wherein the output result is 0 or 1, wherein 0 represents non-atrial fibrillation, and 1 represents atrial fibrillation.
Table 1 the mesh layers are shown in the following table:
effects of the embodiment
1760 electrocardio data known as atrial fibrillation and 2232 electrocardio signals known as non-atrial fibrillation are taken as experimental data, the classification is carried out by using the trained network in the method, and the loss function adopted in the classification is
The results of classification accuracy for different α and β values are shown in table 2:
table 2: accuracy of CNN (CNN) for identifying electrocardio data of atrial fibrillation under different loss functions
It can be seen from the above table that the loss function α -0.8-0.9, β -1.1-1.2, α at 0.8, and β at 1.2, and α at 1.2, and β at 1, achieves the maximum of the non-atrial fibrillation accuracy.
Therefore, as a further improvement, the method for identifying the heart rhythm type based on the improved convolutional neural network of the present application may further perform the following steps:
s1: acquiring an electrocardiosignal to be identified;
s2: preprocessing the electrocardiosignals to obtain electrocardiosignals meeting the identification requirements;
s3: inputting the preprocessed electrocardiosignals into a first neural network trained in advance for recognition;
the pre-trained first neural network uses a loss function as:
wherein c represents the loss size, n represents the number of samples, x represents the samples, y represents the label corresponding to the samples x, a represents the output size calculated by taking the samples x as input and using the current neural network, and α are all regularization coefficients, wherein α is 1-1.2, β is 0.8-1, preferably α is 1.2, and β is 1, and the heart rhythm type corresponding to the electrocardiosignal identified by the first neural network is output;
if the heart rhythm type is non-atrial fibrillation, the electrocardiosignal is considered to be non-atrial fibrillation;
if the heart rhythm type is atrial fibrillation, inputting the preprocessed electrocardiosignals into a pre-trained second neural network again for recognition;
the pre-trained second neural network uses a loss function as:
wherein c denotes a loss magnitude, n denotes a number of samples, x denotes a sample, y denotes a label corresponding to the sample x, a denotes an output magnitude calculated by using the sample x as an input and using the current neural network, and α are all regularization coefficients, wherein α is 0.8-1, β is 1.0-1.2, preferably α is 0.8, and β is 1.2 in the loss function, and the heart rhythm type corresponding to the electrocardiographic signal identified by the second neural network is output, and the heart rhythm type identified by the second neural network is the heart rhythm type of the electrocardiographic signal.
α is 1.2, β is 1, the accuracy rate of recognizing the electrocardiogram of the atrial fibrillation is high, the accuracy rate of recognizing the electrocardiogram of the atrial fibrillation is low, therefore, whether the electrocardiogram is the atrial fibrillation is verified by the first neural network, α is 0.8, β is 1.2, the accuracy rate of recognizing the electrocardiogram of the atrial fibrillation by the second neural network is high, the accuracy rate of recognizing the electrocardiogram of the atrial fibrillation is low, the electrocardiogram of the atrial fibrillation obtained by the first neural network is recognized by the second neural network again, the advantages of the two neural networks can be combined, the accuracy rate of recognition is improved, the electrocardiogram is considered as the type of the atrial fibrillation when the first neural network and the second neural network recognize the electrocardiogram as the type of the atrial fibrillation, and the electrocardiogram is considered as the type of the non-atrial fibrillation when the first neural network and the second neural network recognize the electrocardiogram as the type of the non-atrial fibrillation.
In light of the foregoing description of the preferred embodiments according to the present application, it is to be understood that various changes and modifications may be made without departing from the spirit and scope of the invention. The technical scope of the present application is not limited to the contents of the specification, and must be determined according to the scope of the claims.
As will be appreciated by one skilled in the art, embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the application. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
Claims (8)
1. An apparatus for identifying a heart rhythm type based on an improved convolutional neural network, comprising:
a module for obtaining an electrocardiosignal to be identified;
the module is used for preprocessing the electrocardiosignals to obtain the electrocardiosignals meeting the identification requirement;
the module is used for inputting the preprocessed electrocardiosignals into a first neural network and a second neural network which are trained in advance for recognition;
the pre-trained first neural network uses a loss function as:
wherein c represents a loss magnitude, n represents a number of samples, x represents a sample, y represents a label corresponding to the sample x, a represents an output magnitude calculated by using the sample x as an input and using a current neural network, and α are regularization coefficients, wherein α -1.2 and β -0.8-1 in the loss function output the electrocardiosignals identified by the first neural networkA heart rhythm type;
if the heart rhythm type is non-atrial fibrillation, the electrocardiosignal is considered to be non-atrial fibrillation;
if the heart rhythm type is atrial fibrillation, inputting the preprocessed electrocardiosignals into a pre-trained second neural network again for recognition;
the pre-trained second neural network uses a loss function as:
wherein c represents the loss size, n represents the number of samples, x represents the samples, y represents the label corresponding to the samples x, a represents the output size calculated by taking the samples x as input and using the current neural network, and α are all regularization coefficients, α is 0.8-1, β is 1.0-1.2 in the loss function, and the heart rhythm type corresponding to the electrocardiosignals identified by the second neural network is output;
the electrocardiogram is considered to be of the atrial fibrillation type when the first neural network and the second neural network both identify the electrocardiogram to be of the atrial fibrillation type, and the electrocardiogram is considered to be of the non-atrial fibrillation type when the first neural network and the second neural network identify the electrocardiogram to be of the non-atrial fibrillation type.
2. The apparatus for identifying heart rhythm type based on improved convolution neural network as claimed in claim 1, wherein the first neural network adopts α -1.2 and β -1 in loss function.
3. The apparatus for recognizing rhythm type based on modified convolutional neural network as claimed in claim 1, wherein the second neural network uses α -0.8 and β -1.2 of loss function.
4. The apparatus for identifying heart rhythm type based on improved convolutional neural network of claim 1, wherein the neural network comprises 1 input layer, a plurality of convolutional layers and pooling layers, 1 fully connected layer and 1 classifier layer.
5. The apparatus for identifying heart rhythm type based on improved convolutional neural network as claimed in claim 4, wherein when preprocessing said cardiac electric signal, filter processing is performed on said cardiac electric signal by using a filter with a preset cut-off frequency.
6. The apparatus for recognizing heart rhythm type based on improved convolutional neural network as claimed in claim 5, wherein the preprocessing of the cardiac signal comprises the following steps:
judging whether the sampling frequency of the electrocardiosignals is a preset frequency or not;
and when the sampling frequency is not the preset frequency, resampling the electrocardiosignals into the electrocardiosignals with the preset frequency by adopting an interpolation method.
7. The apparatus for recognizing rhythm type based on improved convolutional neural network as claimed in any of claims 1-6, wherein the pre-trained first and second neural networks are obtained by:
s31: acquiring a training database, wherein the training database is electrocardio data known as atrial fibrillation or non-atrial fibrillation;
s32: preprocessing the electrocardiogram data;
s33: the method is characterized in that a convolutional neural network is used for training, the convolutional neural network is composed of a plurality of convolutional layers, a plurality of pooling layers, a full-link layer and a classifier layer, and known atrial fibrillation or non-atrial fibrillation results are input into the classifier layer during training.
8. The apparatus for identifying heart rhythm type based on improved convolutional neural network of claim 7,
in the step of S31, at least 1 ten thousand of atrial fibrillation electrocardiosignals with 10S and at least 1 ten thousand of other types of electrocardiosignals which are uniformly mixed are used as training data to form a training database, and 0 and 1 are respectively used as labels of the atrial fibrillation electrocardiosignals and the non-atrial fibrillation electrocardiosignals;
in the step of S32, filtering the electrocardio data by using fir filters with upper and lower cut-off frequencies of 0.1Hz and 100Hz respectively, and if the sampling frequency of the electrocardio signals is not 500Hz, resampling the electrocardio signals to 500Hz by using a nearest neighbor interpolation method;
in the step of S33, the convolutional neural network has 8 layers of networks, the 1 st to 6 th layers in the 8 layers of networks are layer1-layer6, and each layer consists of a convolutional layer and a pooling layer, the convolutional layer in the layer1 comprises 5 kernels, the sizes of the convolutional kernels are both 224, and the step sizes and the kernel sizes in the pooling layer in the layer1 are both 2; the layer2 convolutional layer contains 5 kernels, the sizes of the convolutional kernels are all 112, and the step size and the kernel size in the pooling layer in the layer2 are both 2; the layer3 convolutional layer comprises 10 kernels, the sizes of the convolutional kernels are both 100, and the step size and the kernel size in the pooling layer in the layer3 are both 2; the layer4 convolutional layer contains 10 kernels, the sizes of the convolutional kernels are both 50, and the step size and the kernel size in the pooling layer in the layer4 are both 2; the layer5 convolutional layer contains 10 kernels, the sizes of the convolutional kernels are both 48, and the step size and the kernel size in the pooling layer in the layer5 are both 2; the layer6 convolutional layer contains 10 kernels, the sizes of the convolutional kernels are both 24, and the step size and the kernel size in the pooling layer in the layer4 are both 2; the number of output features of layer6 is 30, and finally 10 features are output after calculation of a fully connected layer 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910257087.3A CN110037682B (en) | 2019-04-01 | 2019-04-01 | Method for recognizing heart rhythm type based on improved convolutional neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910257087.3A CN110037682B (en) | 2019-04-01 | 2019-04-01 | Method for recognizing heart rhythm type based on improved convolutional neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110037682A CN110037682A (en) | 2019-07-23 |
| CN110037682B true CN110037682B (en) | 2020-07-21 |
Family
ID=67275777
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910257087.3A Active CN110037682B (en) | 2019-04-01 | 2019-04-01 | Method for recognizing heart rhythm type based on improved convolutional neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110037682B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110037684A (en) * | 2019-04-01 | 2019-07-23 | 上海数创医疗科技有限公司 | Device based on the identification rhythm of the heart type for improving convolutional neural networks |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110751197A (en) * | 2019-10-14 | 2020-02-04 | 上海眼控科技股份有限公司 | Picture classification method, picture model training method and equipment |
| CN111666865B (en) * | 2020-06-02 | 2023-05-23 | 上海数创医疗科技有限公司 | Multi-lead electrocardiosignal convolutional neural network classification method and application method thereof |
| CN112587146B (en) * | 2020-11-25 | 2022-08-16 | 上海数创医疗科技有限公司 | Heart rhythm type identification method and device based on neural network with improved loss function |
| CN117257324B (en) * | 2023-11-22 | 2024-01-30 | 齐鲁工业大学(山东省科学院) | Atrial fibrillation detection method based on convolutional neural network and ECG (electro-magnetic resonance) signals |
Family Cites Families (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106214123B (en) * | 2016-07-20 | 2019-01-25 | 杨一平 | A kind of electrocardiogram compressive classification method based on deep learning algorithm |
| US10709390B2 (en) * | 2017-03-02 | 2020-07-14 | Logos Care, Inc. | Deep learning algorithms for heartbeats detection |
| CN107203692B (en) * | 2017-05-09 | 2020-05-05 | 哈尔滨工业大学（威海） | Electrocardio data digital signal processing method based on deep convolutional neural network |
| CN107316294B (en) * | 2017-06-28 | 2020-03-13 | 太原理工大学 | Lung nodule feature extraction method based on improved depth Boltzmann machine |
| CN108113666B (en) * | 2017-12-19 | 2019-06-21 | 中国科学院深圳先进技术研究院 | Atrial fibrillation signal identification method, device and device |
| CN109063670A (en) * | 2018-08-16 | 2018-12-21 | 大连民族大学 | Block letter language of the Manchus word recognition methods based on prefix grouping |
-
2019
- 2019-04-01 CN CN201910257087.3A patent/CN110037682B/en active Active
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110037684A (en) * | 2019-04-01 | 2019-07-23 | 上海数创医疗科技有限公司 | Device based on the identification rhythm of the heart type for improving convolutional neural networks |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110037682A (en) | 2019-07-23 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110037682B (en) | Method for recognizing heart rhythm type based on improved convolutional neural network | |
| CN110037683B (en) | Improved convolutional neural network for recognizing heart rhythm type and training method thereof | |
| CN110717415B (en) | ST-segment classification convolutional neural network based on feature selection and using method thereof | |
| CN110236520B (en) | Electrocardiogram type recognition device based on double convolution neural network | |
| Cao et al. | Atrial fibrillation detection using an improved multi-scale decomposition enhanced residual convolutional neural network | |
| Ham et al. | Classification of cardiac arrhythmias using fuzzy ARTMAP | |
| CN110717416B (en) | Neural network training method for ST segment classification recognition based on feature selection | |
| CN110037684A (en) | Device based on the identification rhythm of the heart type for improving convolutional neural networks | |
| CN111666865B (en) | Multi-lead electrocardiosignal convolutional neural network classification method and application method thereof | |
| CN104834921B (en) | Electrocardio just/abnormal big data processing method and processing device | |
| CN111956202B (en) | Training method and model of convolutional neural network based on sample distribution improvement | |
| NL2025958B1 (en) | Electrocardiogram signal classification method based on adversarial domain adaptive learning | |
| CN111956208B (en) | A method for ECG signal classification based on ultra-lightweight convolutional neural network | |
| CN111956207A (en) | Electrocardio record marking method, device, equipment and storage medium | |
| CN110367968A (en) | Right bundle branch retardation detection method, device, equipment and storage medium | |
| Hoang et al. | Multi-leads ECG premature ventricular contraction detection using tensor decomposition and convolutional neural network | |
| Park et al. | Cascade classification with adaptive feature extraction for arrhythmia detection | |
| CN108596073A (en) | A kind of lightweight algorithm identifying electrocardiogram (ECG) data based on deep learning | |
| CN112603324B (en) | Neural network training method based on improved loss function | |
| CN110037691B (en) | Improved convolutional neural network for R-wave localization | |
| CN113647959B (en) | Waveform identification method, device and equipment for electrocardiographic waveform signals | |
| CN114519370A (en) | Left ventricular hypertrophy detection method and system based on deep learning early-arrest mechanism | |
| Chakroborty | Accurate Arrhythmia classification using auto-associative neural network | |
| CN111449647B (en) | Electrocardiogram identification method and device, storage medium and electronic equipment | |
| CN117179780A (en) | Electrocardiogram sub-waveform identification method, system, equipment 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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |