CN112426160A - Electrocardiosignal type identification method and device - Google Patents
Electrocardiosignal type identification method and device Download PDFInfo
- Publication number
- CN112426160A CN112426160A CN202011378131.5A CN202011378131A CN112426160A CN 112426160 A CN112426160 A CN 112426160A CN 202011378131 A CN202011378131 A CN 202011378131A CN 112426160 A CN112426160 A CN 112426160A
- Authority
- CN
- China
- Prior art keywords
- standardized
- classification
- sample
- electrocardiosignals
- standard
- 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
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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/253—Fusion techniques of extracted features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Abstract
The invention provides an electrocardiosignal type identification method and device, which train a preset machine learning model by utilizing a classification characteristic sample set marked with an electrocardiosignal type in advance to obtain a classification model, and carry out electrocardiosignal classification processing on time-frequency domain fusion characteristics of standardized electrocardiosignals based on the classification model, thereby improving the identification speed of the electrocardiosignal type and further improving the arrhythmia classification speed. Meanwhile, the classification features input into the classification model are fused with the time domain features and the frequency domain features of the standardized electrocardiosignals, so that the classification features can embody the time domain features of the standardized electrocardiosignals and the frequency domain features of the standardized electrocardiosignals, the accuracy of classification of the electrocardiosignals by the classification model is improved, the quick and accurate identification of the type of the electrocardiosignals is realized, and the quick and accurate identification of the type of arrhythmia is realized.
Description
Technical Field
The invention relates to the technical field of computers, in particular to an electrocardiosignal type identification method and device.
Background
Cardiac arrhythmias are the most common heart diseases and are characterized clinically by irregular cardiac function resulting from disturbances in the rate, rhythm or conduction of the electrical signals of the heart.
At present, doctors generally analyze the duration and amplitude of electrocardiosignal waves according to own experience so as to identify arrhythmia types. However, the method for analyzing the electrocardiosignals based on subjective judgment is time-consuming and may cause signal analysis deviation, and accurate analysis of the electrocardiosignals cannot be realized.
Disclosure of Invention
In view of this, the invention provides a method and a device for identifying the type of an electrocardiographic signal, which classify the electrocardiographic signal based on the time-frequency domain fusion characteristics of the electrocardiographic signal to accurately identify the type of the electrocardiographic signal.
In order to achieve the above purpose, the invention provides the following specific technical scheme:
a method for identifying the type of electrocardiosignals comprises the following steps:
acquiring a standardized electrocardiosignal;
respectively extracting time domain characteristics and frequency domain characteristics of the standardized electrocardiosignals;
performing feature fusion on the time domain features and the frequency domain features to obtain time-frequency domain fusion features of the standardized electrocardiosignals;
inputting the time-frequency domain fusion characteristics into a classification model trained in advance for classification processing to obtain the type of the standardized electrocardiosignals, wherein the type of the standardized electrocardiosignals represents the arrhythmia type corresponding to the standardized electrocardiosignals, and the classification model is obtained by utilizing a classification characteristic sample set with the marked electrocardiosignal type to train a preset machine learning model in advance.
Optionally, the method further includes:
acquiring a history sample set consisting of single-channel electrocardiogram data samples with preset lengths and with electrocardiosignal types marked, and dividing the history sample set into a training set and a testing set according to a preset proportion;
respectively carrying out standardization processing on each sample in the training set and the test set to obtain a standardized parameter and a standard training set and a standard test set;
respectively extracting the time domain characteristics of each sample in the standard training set and the standard testing set by using a long-short term memory network model;
converting each sample in the standard training set and the standard testing set into an image, and respectively extracting the frequency domain characteristics of the corresponding image of each sample in the standard training set and the standard testing set by using a two-dimensional convolutional neural network model;
and respectively carrying out feature fusion on the time domain features and the frequency domain features of each sample in the standard training set and the standard testing set to obtain the classification features of each sample in the standard training set and the standard testing set, thus obtaining a classification feature sample set with the marked electrocardiosignal type.
Optionally, the acquiring the normalized cardiac signal includes:
acquiring original single-channel electrocardiogram data with preset length;
and carrying out standardization processing on the original single-channel electrocardiogram data according to the standardized parameters to obtain the standardized electrocardiogram signals.
Optionally, the respectively extracting the time domain feature and the frequency domain feature of the normalized electrocardiographic signal includes:
extracting the time domain characteristics of the standardized electrocardiosignals by utilizing a pre-trained time domain characteristic extraction model, wherein the time domain characteristic extraction model is obtained by utilizing the standard training set to train a long-short term memory network model in advance, and the standardized electrocardiosignals and samples in the training set have the same format;
and converting the standardized electrocardiosignals into images, inputting the images into a pre-trained frequency domain feature extraction model, and extracting the frequency domain features of the standardized electrocardiosignals, wherein the frequency domain feature extraction model is obtained by utilizing the standard training set to train a two-dimensional convolution neural network model in advance.
Optionally, the method further includes:
training the preset machine learning model by using the classification characteristic and the electrocardiosignal type of each sample in the standard training set;
verifying the classification model obtained after training by using the classification characteristic and the electrocardiosignal type of each sample in the standard test set to obtain a preset performance index value of the classification model;
and when the preset performance index value reaches a threshold value, finishing the training of the classification model.
An apparatus for identifying a type of a cardiac signal, comprising:
the electrocardiosignal acquisition unit is used for acquiring a standardized electrocardiosignal;
the characteristic extraction unit is used for respectively extracting time domain characteristics and frequency domain characteristics of the standardized electrocardiosignals;
the characteristic fusion unit is used for carrying out characteristic fusion on the time domain characteristic and the frequency domain characteristic to obtain a time-frequency domain fusion characteristic of the standardized electrocardiosignal;
and the classification processing unit is used for inputting the time-frequency domain fusion characteristics into a pre-trained classification model for classification processing to obtain the type of the standardized electrocardiosignals, the type of the standardized electrocardiosignals represents the arrhythmia type corresponding to the standardized electrocardiosignals, and the classification model is obtained by utilizing a classification characteristic sample set with the marked electrocardiosignal type to train a preset machine learning model in advance.
Optionally, the apparatus further includes a classification feature sample set obtaining unit, configured to:
acquiring a history sample set consisting of single-channel electrocardiogram data samples with preset lengths and with electrocardiosignal types marked, and dividing the history sample set into a training set and a testing set according to a preset proportion;
respectively carrying out standardization processing on each sample in the training set and the test set to obtain a standardized parameter and a standard training set and a standard test set;
respectively extracting the time domain characteristics of each sample in the standard training set and the standard testing set by using a long-short term memory network model;
converting each sample in the standard training set and the standard testing set into an image, and respectively extracting the frequency domain characteristics of the corresponding image of each sample in the standard training set and the standard testing set by using a two-dimensional convolutional neural network model;
and respectively carrying out feature fusion on the time domain features and the frequency domain features of each sample in the standard training set and the standard testing set to obtain the classification features of each sample in the standard training set and the standard testing set, thus obtaining a classification feature sample set with the marked electrocardiosignal type.
Optionally, the electrocardiograph signal acquiring unit is specifically configured to:
acquiring original single-channel electrocardiogram data with preset length;
and carrying out standardization processing on the original single-channel electrocardiogram data according to the standardized parameters to obtain the standardized electrocardiogram signals.
Optionally, the feature extraction unit is specifically configured to:
extracting the time domain characteristics of the standardized electrocardiosignals by utilizing a pre-trained time domain characteristic extraction model, wherein the time domain characteristic extraction model is obtained by utilizing the standard training set to train a long-short term memory network model in advance, and the standardized electrocardiosignals and samples in the training set have the same format;
and converting the standardized electrocardiosignals into images, inputting the images into a pre-trained frequency domain feature extraction model, and extracting the frequency domain features of the standardized electrocardiosignals, wherein the frequency domain feature extraction model is obtained by utilizing the standard training set to train a two-dimensional convolution neural network model in advance.
Optionally, the apparatus further includes a classification model training unit, specifically configured to:
training the preset machine learning model by using the classification characteristic and the electrocardiosignal type of each sample in the standard training set;
verifying the classification model obtained after training by using the classification characteristic and the electrocardiosignal type of each sample in the standard test set to obtain a preset performance index value of the classification model;
and when the preset performance index value reaches a threshold value, finishing the training of the classification model.
Compared with the prior art, the invention has the following beneficial effects:
the invention discloses an electrocardiosignal type identification method, which is characterized in that a preset machine learning model is trained by utilizing a classification characteristic sample set with marked electrocardiosignal types in advance to obtain a classification model, and electrocardiosignal classification processing is carried out on time-frequency domain fusion characteristics of standardized electrocardiosignals based on the classification model, so that the electrocardiosignal type identification speed is increased, and the arrhythmia classification speed is further increased. Meanwhile, the classification features input into the classification model are fused with the time domain features and the frequency domain features of the standardized electrocardiosignals, so that the classification features can embody the time domain features of the standardized electrocardiosignals and the frequency domain features of the standardized electrocardiosignals, the accuracy of classification of the electrocardiosignals by the classification model is improved, the quick and accurate identification of the type of the electrocardiosignals is realized, and the quick and accurate identification of the type of arrhythmia is realized.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to the provided drawings without creative efforts.
Fig. 1 is a schematic flow chart of a method for identifying an electrocardiographic signal type according to an embodiment of the present invention;
FIG. 2 is a schematic flow chart of a classification model training method according to an embodiment of the present invention;
fig. 3 is a schematic diagram of model processing corresponding to the method for identifying types of electrocardiographic signals disclosed in the embodiment of the present invention;
FIG. 4 is a diagram illustrating a comparison of test results and actual values of a test set according to an embodiment of the present invention;
FIG. 5 is a schematic diagram of evaluation index values of a classification model according to an embodiment of the present invention;
fig. 6 is a schematic structural diagram of an apparatus for recognizing an electrocardiographic signal type according to an embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The inventor finds out through research that: the electrocardiosignals not only have the characteristics of time domain, wherein the bandwidth of the common whole cardiac cycle signals is 0-58 +/-19 Hz, the P wave band is 0-8 +/-3 Hz, the QRS wave bandwidth is 0-55 +/-19 Hz, and the T wave bandwidth is 0-11 +/-2 Hz, so the electrocardiosignals have rich characteristic information in the frequency domain. The electrocardiosignal can be accurately analyzed based on the time domain characteristics and the frequency domain characteristics of the electrocardiosignal, so that the arrhythmia type can be accurately identified. On the basis, the electrocardiosignal is classified based on the time-frequency domain fusion characteristics of the electrocardiosignal, and the arrhythmia type is accurately identified.
Referring to fig. 1, the method for identifying the type of an electrocardiographic signal disclosed in this embodiment includes the following steps:
s101: acquiring a standardized electrocardiosignal;
s102: respectively extracting time domain characteristics and frequency domain characteristics of the standardized electrocardiosignals;
s103: performing feature fusion on the time domain features and the frequency domain features to obtain time-frequency domain fusion features of the standardized electrocardiosignals;
s104: inputting the time-frequency domain fusion characteristics into a classification model trained in advance for classification processing to obtain the type of the standardized electrocardiosignals, wherein the type of the standardized electrocardiosignals represents the arrhythmia type corresponding to the standardized electrocardiosignals, and the classification model is obtained by utilizing a classification characteristic sample set with the marked electrocardiosignal type to train a preset machine learning model in advance.
Specifically, the classification processing result of the classification model is a classification tag value, and the electrocardiograph signal type corresponding to the classification tag value is determined according to the corresponding relationship between the classification tag value and the electrocardiograph signal type, where the electrocardiograph signal type includes: normal (N), left bundle branch block (L), right bundle branch block (R), and ventricular premature beat (V).
It should be noted that, the premise of classifying the electrocardiographic signals by using the classification model is that a preset machine learning model is trained by using a classification feature sample set labeled with the type of the electrocardiographic signals in advance to obtain the classification model.
Referring to fig. 2, the method for training a classification model disclosed in this embodiment includes the following steps:
s201: acquiring a historical sample set consisting of single-channel electrocardiogram data samples with preset lengths and with electrocardiosignal types marked, and dividing the historical sample set into a training set and a testing set according to a preset proportion;
the preset length may be 20 seconds, and the electrocardiographic data is ECG (electrocardiogram) signal data.
The marked electrocardiosignal types comprise: normal (N), left bundle branch block (L), right bundle branch block (R), and ventricular premature beat (V).
The division ratio of the training set and the testing machine can be preset, for example, 20000 samples are divided into training set and testing set according to the ratio of 8:2, wherein the number of samples in the training set and the testing set is 16000 and 4000.
S202: respectively carrying out standardization processing on each sample in the training set and the testing set to obtain a standardized parameter and a standard training set and a standard testing set;
and carrying out standardization processing on each sample in the training set and the test set, unifying the data of each sample in the same data distribution range, and eliminating errors caused by individual differences.
The standardization processing method comprises the following steps: suppose a sequence of variables is x1,…,xnLet us note that the mean and standard deviation are μ and σ, respectively. The normalized sequence is then:
the sequence value after the standardization treatment follows normal distribution, and the standardization parameters mu and sigma are stored for carrying out standardization treatment on the electrocardiogram data to be classified.
The data format after the normalization processing is 1 × 7200 × 1, that is, the data width, height and channel number are 1, 7200 and 1 respectively.
The training set is used for model fitting, and the optimizers of the models all use an Adam method, so that the verification set can not be used in the model training process.
S203: respectively extracting the time domain characteristics of each sample in the standard training set and the standard testing set by using the long-short term memory network model;
fig. 3 shows a schematic diagram of model processing corresponding to the arrhythmia classification method disclosed in this embodiment, where the time domain feature extraction model is LSTM (Long-Short Term Memory), and the activation functions used in the convolutional layer, the full-link layer, and the LSTM Long-Short Term processing time sequence time domain layer in the embodiment of the present invention are ReLU functions, where the corresponding calculation formula is:
s204: converting each sample in the standard training set and the standard testing set into an image, and respectively extracting the frequency domain characteristics of the corresponding image of each sample in the standard training set and the standard testing set by using a two-dimensional convolutional neural network model;
referring to fig. 3, the frequency domain feature extraction model corresponds to a two-dimensional convolutional neural network model, and after converting a sample into an image through time domain transformation, the sample is processed by a convolutional layer 1, a pooling layer 1, a convolutional layer 2, a pooling layer 2, and a full-link layer 1 to obtain frequency domain features.
S205: respectively carrying out feature fusion on the time domain features and the frequency domain features of each sample in the standard training set and the standard testing set to obtain the classification features of each sample in the standard training set and the standard testing set, and obtaining a classification feature sample set with the electrocardiosignal types marked;
referring to fig. 3, the merging layer performs feature fusion on the time domain features and the frequency domain features to obtain classification features.
S206: training a preset machine learning model by using the classification characteristic and the electrocardiosignal type of each sample in the standard training set;
referring to fig. 3, the fully-connected layer 2 is used to map the fused classification features output by the merging layer, the output number of the fully-connected layer is equal to the classification category number of the classification model, and the loss function used for classification is the cross entropy.
S207: verifying the classification model obtained after training by using the classification characteristics and the electrocardiosignal types of each sample in the standard test set to obtain the preset performance index value of the classification model;
optionally, in this embodiment, the control is performed by initializing a set maximum round number (MaxEpochs), and when the number of training iterations reaches the preset maximum round number, the training is ended.
S208: and when the preset performance index value reaches the threshold value, finishing the training of the classification model.
The performance indexes adopted by the embodiment include: accuracy (ACC), sensitivity (Sn), specificity (Sp) and mattescorrelation coefficient (MCC).
The calculation formula of the accuracy (ACC for short) is as follows:
the calculation formula of the sensitivity (Sn for short) is as follows:
wherein the calculation formula of specificity (abbreviated as Sp) is as follows:
wherein the formula for calculating the Martha Correlation Coefficient (MCC) is:
where TP, TN, FP and FN represent examples of true positive, true negative, false positive and false negative, respectively. Values for the madrepore coefficient MCC range from-1 to 1, with an MCC of-1 indicating the worst possible prediction and a value of 1 indicating the best possible prediction scheme. An MCC of 0 indicates random prediction.
The result of comparing the test result of the CNN-LSTM model to the test set and the actual value in this embodiment is obtained by setting the training parameters, as shown in fig. 4. Fig. 4 is a confusion matrix, from which it can be seen that the overall classification accuracy reaches 98.9%, and the classification accuracy of a single class is at least 97.5%.
The evaluation index values shown in fig. 5, including accuracy, sensitivity, specificity and Martha correlation coefficient, can be seen to have good performance for classification of cardiac electrical signals by the CNN-LSTM model.
On the basis, in the practical application of classifying the electrocardiographic signals, in the above embodiment, S101: acquiring a standardized electrocardiosignal, comprising the following steps:
acquiring original single-channel electrocardiogram data with preset length;
according to the standardized parameters obtained by carrying out standardization processing on each sample of the training set test set, the original single-channel electrocardiogram data is subjected to standardization processing to obtain standardized electrocardiosignals, so that the data beyond the range of the training set can be well recovered, and the generalization capability of the model is improved.
S102 in the above embodiment: respectively extracting the time domain characteristics and the frequency domain characteristics of the standardized electrocardiosignals, and the method comprises the following steps:
extracting the time domain characteristics of the standardized electrocardiosignals by utilizing a pre-trained time domain characteristic extraction model, wherein the time domain characteristic extraction model is obtained by utilizing the standard training set to train a long-short term memory network model in advance, and the standardized electrocardiosignals and samples in the training set have the same format;
and converting the standardized electrocardiosignals into images, inputting the images into a pre-trained frequency domain feature extraction model, and extracting the frequency domain features of the standardized electrocardiosignals, wherein the frequency domain feature extraction model is obtained by utilizing the standard training set to train a two-dimensional convolution neural network model in advance.
Therefore, according to the method for identifying the type of the electrocardiosignal disclosed by the embodiment, the preset machine learning model is trained by utilizing the classification feature sample set with the marked electrocardiosignal type in advance to obtain the classification model, and the electrocardiosignal classification processing is performed on the time-frequency domain fusion feature of the standardized electrocardiosignal based on the classification model, so that the identification speed of the type of the electrocardiosignal is increased, and the arrhythmia classification speed is further increased. Meanwhile, the classification features input into the classification model are fused with the time domain features and the frequency domain features of the standardized electrocardiosignals, so that the classification features can embody the time domain features of the standardized electrocardiosignals and the frequency domain features of the standardized electrocardiosignals, the accuracy of classification of the electrocardiosignals by the classification model is improved, the quick and accurate identification of the type of the electrocardiosignals is realized, and the quick and accurate identification of the type of arrhythmia is realized.
Based on the method for identifying the type of the electrocardiographic signal disclosed in the above embodiment, the present embodiment correspondingly discloses an apparatus for identifying the type of the electrocardiographic signal, please refer to fig. 6, and the apparatus for identifying the type of the electrocardiographic signal includes:
an electrocardiographic signal acquisition unit 100 for acquiring a standardized electrocardiographic signal;
a feature extraction unit 200, configured to extract a time domain feature and a frequency domain feature of the normalized electrocardiographic signal, respectively;
a feature fusion unit 300, configured to perform feature fusion on the time domain features and the frequency domain features to obtain time-frequency domain fusion features of the normalized electrocardiographic signal;
the classification processing unit 400 is configured to input the time-frequency domain fusion features into a classification model trained in advance for classification processing, so as to obtain a type of the standardized electrocardiographic signal, where the type of the standardized electrocardiographic signal represents a type of arrhythmia corresponding to the standardized electrocardiographic signal, and the classification model is obtained by training a preset machine learning model in advance by using a classification feature sample set labeled with the type of the electrocardiographic signal.
Optionally, the apparatus further includes a classification feature sample set obtaining unit, configured to:
acquiring a history sample set consisting of single-channel electrocardiogram data samples with preset lengths and with electrocardiosignal types marked, and dividing the history sample set into a training set and a testing set according to a preset proportion;
respectively carrying out standardization processing on each sample in the training set and the test set to obtain a standardized parameter and a standard training set and a standard test set;
respectively extracting the time domain characteristics of each sample in the standard training set and the standard testing set by using a long-short term memory network model;
converting each sample in the standard training set and the standard testing set into an image, and respectively extracting the frequency domain characteristics of the corresponding image of each sample in the standard training set and the standard testing set by using a two-dimensional convolutional neural network model;
and respectively carrying out feature fusion on the time domain features and the frequency domain features of each sample in the standard training set and the standard testing set to obtain the classification features of each sample in the standard training set and the standard testing set, thus obtaining a classification feature sample set with the marked electrocardiosignal type.
Optionally, the electrocardiograph signal acquiring unit 100 is specifically configured to:
acquiring original single-channel electrocardiogram data with preset length;
and carrying out standardization processing on the original single-channel electrocardiogram data according to the standardized parameters to obtain the standardized electrocardiogram signals.
Optionally, the feature extraction unit 200 is specifically configured to:
extracting the time domain characteristics of the standardized electrocardiosignals by utilizing a pre-trained time domain characteristic extraction model, wherein the time domain characteristic extraction model is obtained by utilizing the standard training set to train a long-short term memory network model in advance, and the standardized electrocardiosignals and samples in the training set have the same format;
and converting the standardized electrocardiosignals into images, inputting the images into a pre-trained frequency domain feature extraction model, and extracting the frequency domain features of the standardized electrocardiosignals, wherein the frequency domain feature extraction model is obtained by utilizing the standard training set to train a two-dimensional convolution neural network model in advance.
Optionally, the apparatus further includes a classification model training unit, specifically configured to:
training the preset machine learning model by using the classification characteristic and the electrocardiosignal type of each sample in the standard training set;
verifying the classification model obtained after training by using the classification characteristic and the electrocardiosignal type of each sample in the standard test set to obtain a preset performance index value of the classification model;
and when the preset performance index value reaches a threshold value, finishing the training of the classification model.
According to the device for identifying the type of the electrocardiosignal, the preset machine learning model is trained by utilizing the classification characteristic sample set with the marked electrocardiosignal type in advance to obtain the classification model, and the electrocardiosignal classification processing is carried out on the time-frequency domain fusion characteristic of the standardized electrocardiosignal based on the classification model, so that the identification speed of the type of the electrocardiosignal is increased, and the arrhythmia classification speed is further increased. Meanwhile, the classification features input into the classification model are fused with the time domain features and the frequency domain features of the standardized electrocardiosignals, so that the classification features can embody the time domain features of the standardized electrocardiosignals and the frequency domain features of the standardized electrocardiosignals, the accuracy of classification of the electrocardiosignals by the classification model is improved, the quick and accurate identification of the type of the electrocardiosignals is realized, and the quick and accurate identification of the type of arrhythmia is realized.
The above embodiments can be combined arbitrarily, and the features described in the embodiments in the present specification can be replaced or combined with each other in the above description of the disclosed embodiments, so that those skilled in the art can implement or use the present application.
The previous description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the invention. Thus, the present invention is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims (10)
1. A method for identifying the type of an electrocardiosignal is characterized by comprising the following steps:
acquiring a standardized electrocardiosignal;
respectively extracting time domain characteristics and frequency domain characteristics of the standardized electrocardiosignals;
performing feature fusion on the time domain features and the frequency domain features to obtain time-frequency domain fusion features of the standardized electrocardiosignals;
inputting the time-frequency domain fusion characteristics into a classification model trained in advance for classification processing to obtain the type of the standardized electrocardiosignals, wherein the type of the standardized electrocardiosignals represents the arrhythmia type corresponding to the standardized electrocardiosignals, and the classification model is obtained by utilizing a classification characteristic sample set with the marked electrocardiosignal type to train a preset machine learning model in advance.
2. The method of claim 1, further comprising:
acquiring a history sample set consisting of single-channel electrocardiogram data samples with preset lengths and with electrocardiosignal types marked, and dividing the history sample set into a training set and a testing set according to a preset proportion;
respectively carrying out standardization processing on each sample in the training set and the test set to obtain a standardized parameter and a standard training set and a standard test set;
respectively extracting the time domain characteristics of each sample in the standard training set and the standard testing set by using a long-short term memory network model;
converting each sample in the standard training set and the standard testing set into an image, and respectively extracting the frequency domain characteristics of the corresponding image of each sample in the standard training set and the standard testing set by using a two-dimensional convolutional neural network model;
and respectively carrying out feature fusion on the time domain features and the frequency domain features of each sample in the standard training set and the standard testing set to obtain the classification features of each sample in the standard training set and the standard testing set, thus obtaining a classification feature sample set with the marked electrocardiosignal type.
3. The method of claim 2, wherein said obtaining a normalized cardiac signal comprises:
acquiring original single-channel electrocardiogram data with preset length;
and carrying out standardization processing on the original single-channel electrocardiogram data according to the standardized parameters to obtain the standardized electrocardiogram signals.
4. The method of claim 2, wherein said separately extracting time domain features and frequency domain features of the normalized cardiac electrical signal comprises:
extracting the time domain characteristics of the standardized electrocardiosignals by utilizing a pre-trained time domain characteristic extraction model, wherein the time domain characteristic extraction model is obtained by utilizing the standard training set to train a long-short term memory network model in advance, and the standardized electrocardiosignals and samples in the training set have the same format;
and converting the standardized electrocardiosignals into images, inputting the images into a pre-trained frequency domain feature extraction model, and extracting the frequency domain features of the standardized electrocardiosignals, wherein the frequency domain feature extraction model is obtained by utilizing the standard training set to train a two-dimensional convolution neural network model in advance.
5. The method of claim 2, further comprising:
training the preset machine learning model by using the classification characteristic and the electrocardiosignal type of each sample in the standard training set;
verifying the classification model obtained after training by using the classification characteristic and the electrocardiosignal type of each sample in the standard test set to obtain a preset performance index value of the classification model;
and when the preset performance index value reaches a threshold value, finishing the training of the classification model.
6. An apparatus for identifying a type of an electrocardiographic signal, comprising:
the electrocardiosignal acquisition unit is used for acquiring a standardized electrocardiosignal;
the characteristic extraction unit is used for respectively extracting time domain characteristics and frequency domain characteristics of the standardized electrocardiosignals;
the characteristic fusion unit is used for carrying out characteristic fusion on the time domain characteristic and the frequency domain characteristic to obtain a time-frequency domain fusion characteristic of the standardized electrocardiosignal;
and the classification processing unit is used for inputting the time-frequency domain fusion characteristics into a pre-trained classification model for classification processing to obtain the type of the standardized electrocardiosignals, the type of the standardized electrocardiosignals represents the arrhythmia type corresponding to the standardized electrocardiosignals, and the classification model is obtained by utilizing a classification characteristic sample set with the marked electrocardiosignal type to train a preset machine learning model in advance.
7. The apparatus of claim 6, further comprising a classified feature sample set obtaining unit configured to:
acquiring a history sample set consisting of single-channel electrocardiogram data samples with preset lengths and with electrocardiosignal types marked, and dividing the history sample set into a training set and a testing set according to a preset proportion;
respectively carrying out standardization processing on each sample in the training set and the test set to obtain a standardized parameter and a standard training set and a standard test set;
respectively extracting the time domain characteristics of each sample in the standard training set and the standard testing set by using a long-short term memory network model;
converting each sample in the standard training set and the standard testing set into an image, and respectively extracting the frequency domain characteristics of the corresponding image of each sample in the standard training set and the standard testing set by using a two-dimensional convolutional neural network model;
and respectively carrying out feature fusion on the time domain features and the frequency domain features of each sample in the standard training set and the standard testing set to obtain the classification features of each sample in the standard training set and the standard testing set, thus obtaining a classification feature sample set with the marked electrocardiosignal type.
8. The apparatus according to claim 7, wherein the electrocardiographic signal acquiring unit is specifically configured to:
acquiring original single-channel electrocardiogram data with preset length;
and carrying out standardization processing on the original single-channel electrocardiogram data according to the standardized parameters to obtain the standardized electrocardiogram signals.
9. The apparatus according to claim 7, wherein the feature extraction unit is specifically configured to:
extracting the time domain characteristics of the standardized electrocardiosignals by utilizing a pre-trained time domain characteristic extraction model, wherein the time domain characteristic extraction model is obtained by utilizing the standard training set to train a long-short term memory network model in advance, and the standardized electrocardiosignals and samples in the training set have the same format;
and converting the standardized electrocardiosignals into images, inputting the images into a pre-trained frequency domain feature extraction model, and extracting the frequency domain features of the standardized electrocardiosignals, wherein the frequency domain feature extraction model is obtained by utilizing the standard training set to train a two-dimensional convolution neural network model in advance.
10. The apparatus according to claim 7, further comprising a classification model training unit, specifically configured to:
training the preset machine learning model by using the classification characteristic and the electrocardiosignal type of each sample in the standard training set;
verifying the classification model obtained after training by using the classification characteristic and the electrocardiosignal type of each sample in the standard test set to obtain a preset performance index value of the classification model;
and when the preset performance index value reaches a threshold value, finishing the training of the classification model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011378131.5A CN112426160A (en) | 2020-11-30 | 2020-11-30 | Electrocardiosignal type identification method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011378131.5A CN112426160A (en) | 2020-11-30 | 2020-11-30 | Electrocardiosignal type identification method and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112426160A true CN112426160A (en) | 2021-03-02 |
Family
ID=74699139
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011378131.5A Pending CN112426160A (en) | 2020-11-30 | 2020-11-30 | Electrocardiosignal type identification method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112426160A (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112914585A (en) * | 2021-03-10 | 2021-06-08 | 中山大学 | Single-lead electrocardiogram-oriented classification method, system and device |
| CN113095386A (en) * | 2021-03-31 | 2021-07-09 | 华南师范大学 | Gesture recognition method and system based on three-axis acceleration space-time feature fusion |
| CN113243902A (en) * | 2021-05-31 | 2021-08-13 | 之江实验室 | Feature extraction method based on photoplethysmography |
| CN113598784A (en) * | 2021-08-25 | 2021-11-05 | 济南汇医融工科技有限公司 | Arrhythmia detection method and system |
| CN113688673A (en) * | 2021-07-15 | 2021-11-23 | 电子科技大学 | Cross-user emotion recognition method for electrocardiosignals in online scene |
| CN114259255A (en) * | 2021-12-06 | 2022-04-01 | 深圳信息职业技术学院 | Modal fusion fetal heart rate classification method based on frequency domain signals and time domain signals |
| CN114343665A (en) * | 2021-12-31 | 2022-04-15 | 贵州省人民医院 | Arrhythmia identification method based on graph volume space-time feature fusion selection |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107811626A (en) * | 2017-09-10 | 2018-03-20 | 天津大学 | A kind of arrhythmia classification method based on one-dimensional convolutional neural networks and S-transformation |
| CN108714026A (en) * | 2018-03-27 | 2018-10-30 | 杭州电子科技大学 | The fine granularity electrocardiosignal sorting technique merged based on depth convolutional neural networks and on-line decision |
| CN108852347A (en) * | 2018-07-13 | 2018-11-23 | 京东方科技集团股份有限公司 | For extracting the method for the characteristic parameter of cardiac arrhythmia, the device and computer-readable medium of cardiac arrhythmia for identification |
| CN110432891A (en) * | 2019-07-30 | 2019-11-12 | 天津工业大学 | The feature extraction and classification method of electrocardio beat are extracted in a kind of automation |
| CN110507299A (en) * | 2019-04-11 | 2019-11-29 | 研和智能科技(杭州)有限公司 | Heart rate signal detection device and method |
| CN110742599A (en) * | 2019-11-01 | 2020-02-04 | 广东工业大学 | Electrocardiosignal feature extraction and classification method and system |
| CN111329445A (en) * | 2020-02-20 | 2020-06-26 | 广东工业大学 | Atrial fibrillation identification method based on group convolution residual error network and long-term and short-term memory network |
-
2020
- 2020-11-30 CN CN202011378131.5A patent/CN112426160A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107811626A (en) * | 2017-09-10 | 2018-03-20 | 天津大学 | A kind of arrhythmia classification method based on one-dimensional convolutional neural networks and S-transformation |
| CN108714026A (en) * | 2018-03-27 | 2018-10-30 | 杭州电子科技大学 | The fine granularity electrocardiosignal sorting technique merged based on depth convolutional neural networks and on-line decision |
| CN108852347A (en) * | 2018-07-13 | 2018-11-23 | 京东方科技集团股份有限公司 | For extracting the method for the characteristic parameter of cardiac arrhythmia, the device and computer-readable medium of cardiac arrhythmia for identification |
| CN110507299A (en) * | 2019-04-11 | 2019-11-29 | 研和智能科技(杭州)有限公司 | Heart rate signal detection device and method |
| CN110432891A (en) * | 2019-07-30 | 2019-11-12 | 天津工业大学 | The feature extraction and classification method of electrocardio beat are extracted in a kind of automation |
| CN110742599A (en) * | 2019-11-01 | 2020-02-04 | 广东工业大学 | Electrocardiosignal feature extraction and classification method and system |
| CN111329445A (en) * | 2020-02-20 | 2020-06-26 | 广东工业大学 | Atrial fibrillation identification method based on group convolution residual error network and long-term and short-term memory network |
Non-Patent Citations (2)
| Title |
|---|
| B. MURUGESAN等: "ECGNet: Deep Network for Arrhythmia Classification", 《2018 IEEE INTERNATIONAL SYMPOSIUM ON MEDICAL MEASUREMENTS AND APPLICATIONS(MEMEA)》 * |
| F.LIU等: "A LSTM AND CNN BASED ASSEMBLE NEURAL NETWORK FRAMEWORK FOR", 《2019 IEEE INTERNATIONAL CONFERENCE ON ACOUSTICS,SPEECH AND SIGNAL PROCESSING》 * |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112914585A (en) * | 2021-03-10 | 2021-06-08 | 中山大学 | Single-lead electrocardiogram-oriented classification method, system and device |
| CN113095386A (en) * | 2021-03-31 | 2021-07-09 | 华南师范大学 | Gesture recognition method and system based on three-axis acceleration space-time feature fusion |
| CN113095386B (en) * | 2021-03-31 | 2023-10-13 | 华南师范大学 | Gesture recognition method and system based on triaxial acceleration space-time feature fusion |
| CN113243902A (en) * | 2021-05-31 | 2021-08-13 | 之江实验室 | Feature extraction method based on photoplethysmography |
| CN113688673A (en) * | 2021-07-15 | 2021-11-23 | 电子科技大学 | Cross-user emotion recognition method for electrocardiosignals in online scene |
| CN113688673B (en) * | 2021-07-15 | 2023-05-30 | 电子科技大学 | Cross-user emotion recognition method for electrocardiosignals in online scene |
| CN113598784A (en) * | 2021-08-25 | 2021-11-05 | 济南汇医融工科技有限公司 | Arrhythmia detection method and system |
| CN113598784B (en) * | 2021-08-25 | 2024-04-09 | 济南汇医融工科技有限公司 | Arrhythmia detection method and system |
| CN114259255A (en) * | 2021-12-06 | 2022-04-01 | 深圳信息职业技术学院 | Modal fusion fetal heart rate classification method based on frequency domain signals and time domain signals |
| CN114259255B (en) * | 2021-12-06 | 2023-12-08 | 深圳信息职业技术学院 | Modal fusion fetal heart rate classification method based on frequency domain signals and time domain signals |
| CN114343665A (en) * | 2021-12-31 | 2022-04-15 | 贵州省人民医院 | Arrhythmia identification method based on graph volume space-time feature fusion selection |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112426160A (en) | Electrocardiosignal type identification method and device | |
| CN108186011B (en) | Atrial fibrillation detection method, atrial fibrillation detection device and readable storage medium | |
| CN107822622B (en) | Electrocardiogram diagnosis method and system based on deep convolutional neural network | |
| CN111160139B (en) | Electrocardiosignal processing method and device and terminal equipment | |
| KR102451795B1 (en) | ECG signal detection method | |
| CN109276242A (en) | The method and apparatus of electrocardiosignal type identification | |
| CN111345816B (en) | Multi-lead QRS complex detection method, device, equipment and storage medium | |
| CN111345817B (en) | QRS complex position determination method, device, equipment and storage medium | |
| CN113057648A (en) | ECG signal classification method based on composite LSTM structure | |
| CN108742697B (en) | Heart sound signal classification method and terminal equipment | |
| CN113995419B (en) | Atrial fibrillation risk prediction system based on heartbeat rhythm signal and application thereof | |
| CN110693488A (en) | Electrocardiosignal processing system, electrocardiosignal processing method, electronic equipment and readable storage medium | |
| CN112257518A (en) | ECG identity recognition method based on WT and WOA-PNN algorithm | |
| Rohmantri et al. | Arrhythmia classification using 2D convolutional neural network | |
| CN110897634A (en) | Electrocardiosignal generation method based on generation countermeasure network | |
| CN111956214A (en) | QRS wave automatic detection method based on U-net end-to-end neural network model | |
| CN110558975A (en) | Electrocardiosignal classification method and system based on LRF-ELM and BLSTM | |
| CN111053552A (en) | QRS wave detection method based on deep learning | |
| CN112883803B (en) | Electrocardiogram signal classification method, device and storage medium based on deep learning | |
| CN113343869A (en) | Electroencephalogram signal automatic classification and identification method based on NTFT and CNN | |
| CN115486849A (en) | Electrocardiosignal quality evaluation method, device and equipment | |
| CN113288157A (en) | Arrhythmia classification method based on depth separable convolution and improved loss function | |
| CN116712099A (en) | Fetal heart state detection method based on multi-mode data, electronic equipment and storage medium | |
| Boonyakitanont et al. | ScoreNet: a Neural network-based post-processing model for identifying epileptic seizure onset and offset in EEGs | |
| CN113647959B (en) | Waveform identification method, device and equipment for electrocardiographic waveform signals |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20210302 |
|
| RJ01 | Rejection of invention patent application after publication |