CN111067512B - Ventricular fibrillation detection device, ventricular fibrillation detection model training method and equipment - Google Patents
Ventricular fibrillation detection device, ventricular fibrillation detection model training method and equipment Download PDFInfo
- Publication number
- CN111067512B CN111067512B CN201911237611.7A CN201911237611A CN111067512B CN 111067512 B CN111067512 B CN 111067512B CN 201911237611 A CN201911237611 A CN 201911237611A CN 111067512 B CN111067512 B CN 111067512B
- Authority
- CN
- China
- Prior art keywords
- ventricular fibrillation
- data
- heart
- heart beat
- training
- 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]
- 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
- A61B5/02—Detecting, measuring or recording pulse, heart rate, blood pressure or blood flow; Combined pulse/heart-rate/blood pressure determination; Evaluating a cardiovascular condition not otherwise provided for, e.g. using combinations of techniques provided for in this group with electrocardiography or electroauscultation; Heart catheters for measuring blood pressure
- A61B5/024—Detecting, measuring or recording pulse rate or heart rate
- A61B5/02405—Determining heart rate variability
-
- 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
Landscapes
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Cardiology (AREA)
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Animal Behavior & Ethology (AREA)
- Public Health (AREA)
- Biomedical Technology (AREA)
- Heart & Thoracic Surgery (AREA)
- Medical Informatics (AREA)
- Molecular Biology (AREA)
- Surgery (AREA)
- Biophysics (AREA)
- General Health & Medical Sciences (AREA)
- Pathology (AREA)
- Veterinary Medicine (AREA)
- Physiology (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- Fuzzy Systems (AREA)
- Mathematical Physics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Psychiatry (AREA)
- Signal Processing (AREA)
- Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)
Abstract
The application discloses ventricular fibrillation detection device, training device, method and equipment of detection model, the detection device includes: the first extraction unit is used for extracting a preset number of heartbeat data from the heart rhythm data to be detected to form a heartbeat set; the second extraction unit is used for performing feature extraction on the heartbeat set by using the heartbeat feature extraction model to obtain heartbeat features; the heart beat feature extraction model is obtained by training a pre-training model by using ventricular fibrillation data samples; and the detection unit is used for inputting the heart beat characteristics and the heart rhythm characteristics into the trained ventricular fibrillation detection model to obtain a detection result. The method and the device have the advantages that the arrhythmia data samples with large quantity are used for pre-training the heartbeat feature extraction model, so that the training precision of the heartbeat feature extraction model can be guaranteed under the condition that ventricular fibrillation data samples are insufficient. And furthermore, ventricular fibrillation detection is performed on the heart rhythm data to be detected based on the heart beat features extracted by the heart beat feature extraction model, so that the obtained detection result is more accurate.
Description
Technical Field
The application relates to the field of data processing, in particular to a ventricular fibrillation detection device, a model detection training device, a model detection method and related equipment.
Background
Ventricular Fibrillation (VF), which is called ventricular fibrillation for short, is a serious type of arrhythmia, and refers to a fatal arrhythmia in which unordered activation of ventricles causes regular and ordered activation and contraction of ventricles to disappear. Clinical symptoms include sudden loss of consciousness, convulsions, respiratory arrest and even death, loss of auscultatory heart sounds, inability to palpate pulse, and inability to detect blood pressure. The self-recovery is very little, rapid cardio-pulmonary-cerebral resuscitation is often needed, and after more than 10 minutes, the cardio-pulmonary-cerebral resuscitation is difficult to succeed.
At present, the detection of the ventricular fibrillation is usually realized based on a deep learning model, but because the number of ventricular fibrillation data samples is small, the training precision for training the deep learning model based on the training samples with insufficient number cannot be guaranteed, and further, the result of detecting the ventricular fibrillation by using the trained deep learning model is inaccurate.
Disclosure of Invention
In view of this, the present application provides a detecting device for ventricular fibrillation, a training device for a model, a method for the detecting device, and related equipment, which can pre-train a heartbeat feature extraction model by using a large number of arrhythmia data samples, so that the training precision of the model can be ensured even when the ventricular fibrillation data samples are insufficient, and the accuracy of the ventricular fibrillation detection result is further improved.
In a first aspect, to achieve the above object, the present application provides a ventricular fibrillation detection apparatus, including:
the first extraction unit is used for extracting a preset number of heartbeat data from the heart rate data to be detected to form a heartbeat set of the heart rate data to be detected; the heart rate data to be detected comprises a plurality of heart beat cycles, and each heart beat data comprises one heart beat cycle;
the second extraction unit is used for performing feature extraction on the heart beat set by using a heart beat feature extraction model to obtain heart beat features of the heart beat data to be detected; the heart beat feature extraction model is obtained by training a pre-training model by using ventricular fibrillation data samples, and the pre-training model is obtained by training arrhythmia data samples;
and the detection unit is used for inputting the heartbeat characteristics and the heart rhythm characteristics extracted in advance into a trained ventricular fibrillation detection model to obtain a detection result of the heart rhythm data to be detected.
In an alternative embodiment, the apparatus further comprises:
the third extraction unit is used for extracting the HRV (heart rate variability) characteristics of the heart rate data to be detected; wherein the HRV characteristic belongs to the heart rhythm characteristic.
In an alternative embodiment, the apparatus further comprises:
the fourth extraction unit is used for extracting IMF characteristics of the heart rhythm data to be detected; wherein the IMF characteristic belongs to the heart rhythm characteristic.
In an optional implementation, the fourth extraction unit includes:
the first extraction subunit is used for extracting time domain IMF characteristics of the heart rhythm data to be detected;
and/or the second extraction subunit is used for extracting the frequency domain IMF characteristics of the heart rhythm data to be detected.
In an alternative embodiment, the first extraction subunit includes:
the third extraction subunit is used for extracting an IMF1 signal and a residual signal of the heart rhythm data to be detected;
and the first calculating subunit is used for calculating the similarity between the IMF1 signal and the residual signal and the to-be-detected rhythm data respectively in a time domain range, and the similarity is used as the time domain IMF characteristics of the to-be-detected rhythm data.
In an optional embodiment, the second extraction subunit includes:
the fourth extraction subunit is used for extracting the IMF1 signal and the residual signal of the heart rhythm data to be detected;
and the second calculating subunit is used for calculating the similarity between the IMF1 signal and the residual signal and the data of the heart rhythm to be detected respectively in a frequency domain range, and taking the similarity as the frequency domain IMF characteristics of the data of the heart rhythm to be detected.
In a second aspect, the present application provides a training apparatus for ventricular fibrillation detection models, which is used in the ventricular fibrillation detection apparatus described in any one of the above embodiments, and the apparatus includes:
the first extraction module is used for extracting a preset number of heartbeat data from the ventricular fibrillation data samples to form a heartbeat set of the ventricular fibrillation data samples; wherein the ventricular fibrillation data samples comprise a plurality of heart beat cycles, each heart beat data sample comprising a heart beat cycle;
the second extraction module is used for performing feature extraction on the heart beat set by using a heart beat feature extraction model to obtain heart beat features of the ventricular fibrillation data sample; the heart beat feature extraction model is obtained by utilizing arrhythmia samples for pre-training;
and the training module is used for training the ventricular fibrillation detection model based on the heart beat characteristics of the ventricular fibrillation data samples and the heart rhythm characteristics extracted in advance to obtain the trained ventricular fibrillation detection model.
In an alternative embodiment, the apparatus further comprises:
the third extraction module is used for extracting HRV (high resolution video) features of the ventricular fibrillation data samples; wherein the HRV characteristic belongs to the heart rhythm characteristic.
In an alternative embodiment, the apparatus further comprises:
the fourth extraction module is used for extracting IMF characteristics of the ventricular fibrillation data samples; wherein the IMF characteristic belongs to the heart rhythm characteristic.
In a third aspect, the present application provides a method for training a ventricular fibrillation detection model, where the ventricular fibrillation detection model is used in any one of the above ventricular fibrillation detection devices, and the method includes:
extracting a preset number of heartbeat data from ventricular fibrillation data samples to form a heartbeat set of the ventricular fibrillation data samples; wherein the ventricular fibrillation data samples comprise a plurality of heart beat cycles, each heart beat data sample comprising one heart beat cycle;
performing feature extraction on the heart beat set by using a heart beat feature extraction model to obtain heart beat features of the ventricular fibrillation data sample; the heart beat feature extraction model is obtained by utilizing arrhythmia samples for pre-training;
and training the ventricular fibrillation detection model based on the heart beat features of the ventricular fibrillation data samples and the heart rhythm features extracted in advance to obtain the trained ventricular fibrillation detection model.
In an optional embodiment, the method further comprises:
extracting HRV characteristics of the ventricular fibrillation data samples; wherein the HRV characteristic belongs to the heart rhythm characteristic.
In an optional embodiment, the method further comprises:
extracting IMF characteristics of the ventricular fibrillation data sample; wherein the IMF characteristic belongs to the heart rhythm characteristic.
In a fourth aspect, the present application further provides a computer-readable storage medium having stored therein instructions, which when run on a terminal device, cause the terminal device to implement the functions of the apparatus or perform the method of any of the preceding claims.
In a fifth aspect, the present application further provides an apparatus, comprising: memory, a processor, and a computer program stored on the memory and executable on the processor, when executing the computer program, implementing the functions of the apparatus of any of the preceding claims or performing the method of any of the preceding claims.
In the ventricular fibrillation detection device provided by the embodiment of the application, the cardiac beat feature extraction model is pre-trained by using a large number of arrhythmia data samples, so that the training precision of the cardiac beat feature extraction model can be ensured under the condition that the ventricular fibrillation data samples are insufficient. And furthermore, ventricular fibrillation detection is performed on the heart rhythm data to be detected based on the heart beat features and the heart rhythm features extracted by the heart beat feature extraction model, so that the obtained detection result is more accurate.
Further, this application still provides a training method and device of ventricular fibrillation detection model, based on heart beat characteristic and the rhythm of the heart characteristic that above-mentioned heart beat characteristic extraction model extracted, trains ventricular fibrillation detection model, can obtain more accurate ventricular fibrillation detection model to improve the accuracy that ventricular fibrillation detected.
Drawings
In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required to be used in the description of the embodiments are briefly introduced below, and it is obvious that the drawings in the description below are only some embodiments of the present application, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.
Fig. 1 is a schematic structural diagram of a ventricular fibrillation detection apparatus according to an embodiment of the present disclosure;
fig. 2 is a schematic structural diagram of a pre-training model constructed based on a ResNet convolutional neural network according to an embodiment of the present disclosure;
FIG. 3 is a schematic waveform diagram provided in accordance with an embodiment of the present application;
fig. 4 is a schematic structural diagram of a training device of a ventricular fibrillation detection model according to an embodiment of the present application;
FIG. 5 is a flowchart of a training method for a ventricular fibrillation detection model according to the embodiment of the present application;
FIG. 6 is a block diagram of a ventricular fibrillation detection apparatus according to an embodiment of the present application;
fig. 7 is a structural diagram of a training apparatus of a ventricular fibrillation detection model according to an embodiment of the present application.
Detailed Description
The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are only a part of the embodiments of the present application, and not all of the embodiments. All other embodiments obtained by a person of ordinary skill in the art based on the embodiments in the present application without making any creative effort belong to the protection scope of the present application.
In the existing ventricular fibrillation detection method, due to the limitation of the number of ventricular fibrillation data samples, the precision of a heart beat feature extraction model is low, the precision of a ventricular fibrillation detection model trained based on heart beat features extracted by the heart beat extraction model is further low, and finally the accuracy of the result of ventricular fibrillation detection by using the ventricular fibrillation detection model is low.
Therefore, the ventricular fibrillation detection device provided by the application utilizes a large number of arrhythmia data samples to pre-train the heartbeat feature extraction model, so that the training precision of the heartbeat feature extraction model can be ensured under the condition that the ventricular fibrillation data samples are insufficient. And furthermore, ventricular fibrillation detection is performed on the heart rhythm data to be detected based on the heart beat features extracted by the heart beat feature extraction model, so that the obtained detection result is more accurate.
The present application provides a ventricular fibrillation detection apparatus, and with reference to fig. 1, a schematic structural diagram of a ventricular fibrillation detection apparatus provided in an embodiment of the present application, where the apparatus includes:
the first extraction unit 101 is configured to extract a preset number of heartbeat data from heart rate data to be detected to form a heartbeat set of the heart rate data to be detected; the heart rate data to be detected comprises a plurality of heart beat cycles, and each heart beat data comprises one heart beat cycle;
the second extraction unit 102 is configured to perform feature extraction on the heartbeat set by using a heartbeat feature extraction model to obtain heartbeat features of the heart rhythm data to be detected; the heart beat feature extraction model is obtained by training a pre-training model by using ventricular fibrillation data samples, and the pre-training model is obtained by training arrhythmia data samples;
and the detection unit 103 is used for inputting the heart beat features and the heart rhythm features extracted in advance into the trained ventricular fibrillation detection model to obtain a detection result of the heart rhythm data to be detected.
In the first extraction unit 101, the heart rhythm data to be detected is electrocardiographic data including a plurality of heart cycles, and is electrocardiographic signals obtained by monitoring the heart, and by analyzing the heart rhythm data to be detected, a corresponding heart beating rule can be obtained, and the heart rhythm data can be used as auxiliary information for diagnosing various cardiovascular and cerebrovascular diseases.
In the embodiment of the application, after a section of heart rhythm data to be detected is acquired, a preset number of heart beat data are extracted from the heart rhythm data to be detected, and the heart beat data are used for forming a heart beat set of the heart rhythm data to be detected. Specifically, there are many ways to extract heartbeat data, which is not limited in the embodiments of the present application.
In an alternative embodiment, heartbeat data can be extracted from the heart rhythm data to be detected by the R-wave position. Specifically, firstly, the position of each R wave in the data of the heart rhythm to be detected is determined, and secondly, the complete heart cycle including the position of one R wave in the data of the heart rhythm to be detected is determined as the heart rhythm data of the heart rhythm to be detected. Usually, one heart rate data to be detected contains a plurality of heart beat cycles, that is, positions with a plurality of R-waves, and therefore, a plurality of heart beat data can be extracted from the heart rate data to be detected.
In practical application, after the position of any R wave in the heart rhythm data to be detected is determined, the intervals between two adjacent R waves taking the position of the R wave as the center and the R wave at the center are determined respectively. In an alternative embodiment, a 1/3 length segment of the interval between the center R-wave and the adjacent left R-wave and a 2/3 length segment of the interval between the center R-wave and the adjacent right R-wave may be truncated to collectively form heartbeat data for one complete heartbeat cycle including the center R-wave.
In the embodiment of the application, a plurality of heart cycles can be extracted from the heart rhythm data to be detected according to the above manner, so as to form a heart beat set of the heart rhythm data to be detected, and the heart beat set is used for extracting heart beat features of the heart rhythm data to be detected subsequently.
In the second extraction unit 102, because the sample size of the ventricular fibrillation data sample is small, the heartbeat feature extraction model is completely trained based on the ventricular fibrillation data sample, and obviously, the training precision of the heartbeat feature extraction model cannot be ensured. Therefore, according to the embodiment of the application, the arrhythmia data samples similar to the ventricular fibrillation data samples are used for pre-training the heartbeat feature extraction model to obtain the pre-training model, then the ventricular fibrillation data samples with small sample size are used for further training the pre-training model to obtain the trained heartbeat feature extraction model, and the training precision of the heartbeat feature extraction model can be obviously improved.
The arrhythmia data samples may include normal heartbeat (labeled N) data samples, left bundle branch block (labeled L) data samples, right bundle branch block (labeled R) data samples, atrial premature (labeled a) data samples, ventricular premature (labeled P) data samples, and ventricular fibrillation (labeled V) data samples.
In order to ensure the distribution balance of various types of samples in arrhythmia data samples, so that the deep learning network is easy to converge and optimize, the large number of normal heart beat data samples in the arrhythmia data samples can be randomly eliminated.
In an optional implementation manner, a threshold value of 0.2 may be set, when a data sample labeled as a normal heart beat N is detected in a heart beat abnormality data sample, a random number of 0 to 1 is generated, if the random number is smaller than the threshold value of 0.2, the data sample is retained as an arrhythmia data sample, otherwise, the data sample is deleted from the arrhythmia data sample and is not used in the pre-training of the heart beat feature extraction model, so that the proportion balance of the data samples of various heart beat types in the arrhythmia data sample used for the pre-training of the heart beat feature extraction model is realized, and the accuracy of the pre-training of the heart beat feature extraction model is further improved.
Before pre-training the cardiac rhythm feature extraction model by using the cardiac arrhythmia data samples, the cardiac arrhythmia data samples are first processed, specifically, a preset number of cardiac rhythm data are extracted from each cardiac arrhythmia data sample, and the specific extraction mode may refer to a mode of extracting a preset number of cardiac rhythm data from cardiac rhythm data to be detected in the first extraction unit 101, which is not described herein again. It should be noted that, when determining the position of the R-wave in the arrhythmia data sample, the position of the R-wave may be determined based on the characteristics of various types of arrhythmia data samples, and details thereof are not repeated here. In addition, the label of the arrhythmia data sample serves as a label of the corresponding heartbeat data.
In the embodiment of the application, after cardiac beat data with labels are extracted from arrhythmia data samples, the cardiac beat feature extraction model is pre-trained by using the extracted cardiac beat data and the corresponding labels to obtain a pre-training model.
In the embodiment of the application, a pre-training model trained by using a large number of arrhythmia data samples is migrated to a heart beat feature extraction model based on a transfer learning method, ventricular fibrillation data samples are further trained on the basis of the pre-training model, the trained heart beat feature extraction model is finally obtained, and the training precision of the heart beat feature extraction model obtained by the transfer learning method can be ensured.
The ventricular fibrillation data samples include normal heartbeat (labeled N) data samples and ventricular fibrillation (labeled V) data samples, typically from a cardiac arrest data set, and are small in number.
Before the pre-training model is further trained by using the ventricular fibrillation data samples, the ventricular fibrillation data samples are processed, specifically, a preset number of heartbeat data are extracted from each ventricular fibrillation data sample, and the specific extraction mode may refer to a mode of extracting a preset number of heartbeat data from the heart rhythm data to be detected in the first extraction unit 101, which is not described herein again. It should be noted that, when determining the position of the R wave in the ventricular fibrillation data sample, the position of the R wave may be determined based on the feature corresponding to the label of the ventricular fibrillation data sample, which is not described herein again. In addition, the label of the ventricular fibrillation data sample serves as a label of the corresponding heartbeat data.
In the embodiment of the application, after heartbeat data with labels are extracted from a ventricular fibrillation data sample, the pre-training model is further trained by using the extracted heartbeat data and the corresponding labels, and a trained heartbeat feature extraction model is obtained and used for extracting heartbeat features of heart rhythm data to be detected.
In practical application, after the first extraction unit 101 extracts the heartbeat set of the heart rate data to be detected, the second extraction unit 102 performs feature extraction on the heartbeat set by using the trained heartbeat feature extraction model to obtain heartbeat features of the heart rate data to be detected, and the heartbeat features are used for detecting the heart rate data to be detected.
For a pre-training model for transfer learning to a heart beat feature extraction model, an embodiment of the present application provides a construction method based on a ResNet convolutional neural network, and as shown in fig. 2, a schematic structural diagram of the pre-training model constructed based on the ResNet convolutional neural network is provided in the embodiment of the present application. Specifically, input layer (Input): the input parameters are one-dimensional heartbeat data subjected to preprocessing (such as filtering processing and the like), for example, the data shape is 432 × 1, and a batch processing mode can be adopted, so that 1024 pieces of heartbeat data can be input by the input layer at a time. Convolutional layer (Conv): one-dimensional convolution can be adopted, wherein the length of a convolution kernel can be 5, the number of input channels can be 1, and the number of output channels can be 32.BN layer: and adopting a batch labeling layer for recording the mean value and the variance of each layer of the pre-training model in the pre-training process, wherein the attenuation coefficient can be 0.997. Activation layer (ReLu): a non-linear activation function ReLu may be employed as the activation function of the pre-trained model. Residual Block (Residual Block): the length of a convolution kernel in the convolution layer can be 5, the input and output channels can be 32, the residual error part can adopt a summation mode, and the weight and the offset are set to be 0, so that f (a) =0, the depth of a network is ensured, and the phenomenon of gradient explosion or gradient dispersion caused by coefficient transmission due to overlarge network can be avoided; the Pooling layer may use a Max Pooling method (Max Pooling) where the scale may be 1 and the sliding distance may be 1, and the residual block may be stacked 2 times in a model construction. Fully connected layer (Dense): the number of the neurons output by the full connection layer can be 32, and the conclusion of the layer can be stored in the pre-training model to be used as an extraction layer of transfer learning. Classification layer (Softmax): for outputting a conclusion for each arrhythmia data sample, the above 6 (N, L, R, a, P, V) types can be classified.
In addition, in the model construction mode, adam optimizer is selected to optimize the model parameters, and in a specific implementation mode, the selected step length is 0.001, the moment estimation attenuation rates are ρ 1=0.9 and ρ 2=0.999 respectively, the batch size is set to 1024, the loss function is cross entropy, the iteration cycle is 10000 times, and the like. It should be noted that the manner of optimizing the model and the manner of setting the parameters are not intended to limit the embodiments of the present application.
In practical application, in the stage of pre-training by using the arrhythmia data model, learning parameters of the pre-training model are obtained, and in order to realize transfer learning, the learning parameters obtained in the pre-training stage are stored in the embodiment of the application, so that after the pre-training model is subjected to learning transfer, the heart beat features for detecting ventricular fibrillation can be extracted by using the last hidden layer of the pre-training model.
In the detection unit 103, because the heart rhythm data to be detected has not only heart beat characteristics but also heart rhythm characteristics, in order to improve the accuracy of the detection result of the ventricular fibrillation, the embodiment of the present application may detect the heart rhythm data to be detected by combining the heart beat characteristics and the heart rhythm characteristics of the heart rhythm data to be detected.
The HRV feature is used for reflecting the correlation of the heartbeat interval difference of successive heartbeats in the electrocardio data and essentially analyzing the heart cycle variability. In the embodiment of the application, the HRV feature is taken as one of the heart rhythm features and is used for detecting ventricular fibrillation of the heart rhythm data to be detected. The HRV feature extraction method is not limited in the embodiments of the present application.
IMF characteristics for describing oscillation characteristics of the electrocardiosignal. The inventor discovered a phenomenon in the research, as shown in fig. 3, the left side is the waveform of ventricular fibrillation signal (VF), wherein the IMF1 signal of ventricular fibrillation signal is similar to the filtered waveform of original ventricular fibrillation signal, and the residual signal R of ventricular fibrillation signal is not consistent with the filtered waveform of original ventricular fibrillation signal. And the right is the waveform of the non-ventricular fibrillation signal (not VF), wherein the IMF1 signal of the non-ventricular fibrillation signal is inconsistent with the filtered waveform of the original ventricular fibrillation signal, and the residual signal R of the ventricular fibrillation signal is similar to the filtered waveform of the original ventricular fibrillation signal, so that the conclusion of the ventricular fibrillation signal is exactly opposite to that of the non-ventricular fibrillation signal. Therefore, a method of representing the IMF characteristics may be determined based on the above conclusions that can distinguish ventricular fibrillation signals from non-ventricular fibrillation signals.
In an optional implementation manner, in the embodiment of the present application, firstly, an IMF1 signal and a residual signal R of rhythm data to be detected are extracted, and secondly, in a time domain range, similarities between the IMF1 signal and the residual signal and the rhythm data to be detected, respectively, are calculated, and are used as time domain IMF features of the rhythm data to be detected. Specifically, the similarity between the IMF1 signal and the residual signal and the data of the heart rhythm to be detected can be calculated in a time domain range by using formulas (1) and (2):
wherein, signal is used for representing the data of the heart rhythm to be detected, IMF1 is used for representing MF1 signals of the data of the heart rhythm to be detected, and R is used for representing residual signals of the data of the heart rhythm to be detected.
In order to prevent interference signals of some unknown frequency factors and increase the robustness of the algorithm, the similarity between the IMF1 signal and the residual signal and the heart rhythm data to be detected can be calculated in the frequency domain range to serve as the frequency domain IMF characteristics of the heart rhythm data to be detected. Specifically, the similarity between the IMF1 signal and the residual signal and the data of the heart rhythm to be detected can be calculated in a frequency domain range by using the following formulas (3) and (4):
the Signal (DFT) is used for representing frequency domain signals of the heart rate data to be detected, the IMF1 (DFT) is used for representing MF1 signals of the frequency domain signals of the heart rate data to be detected, and the R is used for representing residual signals of the frequency domain signals of the heart rate data to be detected.
For the heart beat features, in an optional implementation manner, the number of neurons output by the full connection layer of the heart beat feature extraction model is 32, and the heart beat data to be detected contains 10 heart beat cycles, that is, 10 heart beat data, and the heart beat features of the heart beat data to be detected extracted by the heart beat feature extraction model are 32 × 10. Because the number of features of heartbeat features is large, in order to improve the processing efficiency, the embodiment of the application can adopt a main program analysis method (PCA) and the like to perform dimensionality reduction processing on the heartbeat features. Specifically, the heart beat features corresponding to the heart beat set extracted by using the heart beat feature extraction model are merged, for example, by means of mean processing, covariance feature values, feature vector solving and the like, the heart beat features of the heart beat data to be detected after dimensionality reduction are finally obtained. It should be noted that the embodiment of the present application does not limit the dimensionality reduction mode of the heartbeat feature.
In the detection unit 103, after the heart beat features, the HRV features, the time domain IMF features, and the frequency domain IMF features of the heart rate data to be detected are obtained, the heart beat features, the HRV features, the time domain IMF features, and the frequency domain IMF features are input into the trained ventricular fibrillation detection model, so that the heart rate data to be detected is detected, and a detection result is obtained. Specifically, the ventricular fibrillation detection model is obtained by training a ventricular fibrillation data sample, and the detection result belongs to a ventricular fibrillation type or a normal type.
In the ventricular fibrillation detection device provided by the embodiment of the application, the heartbeat feature extraction model is pre-trained by using a large number of arrhythmia data samples, so that the training precision of the heartbeat feature extraction model can be ensured under the condition that the ventricular fibrillation data samples are insufficient. And furthermore, ventricular fibrillation detection is performed on the heart rhythm data to be detected based on the heart beat features and the heart rhythm features extracted by the heart beat feature extraction model, so that the obtained detection result is more accurate.
Based on the foregoing embodiment, an embodiment of the present application further provides a training apparatus for a ventricular fibrillation detection model, where the ventricular fibrillation detection model is used in the aforementioned ventricular fibrillation detection apparatus, see fig. 4, and is a structural schematic diagram of the training apparatus for the ventricular fibrillation detection model provided in the embodiment of the present application, where the training apparatus includes:
the first extraction module 401 is configured to extract a preset number of heartbeat data from ventricular fibrillation data samples to form a heartbeat set of the ventricular fibrillation data samples; wherein the ventricular fibrillation data samples comprise a plurality of heart beat cycles, each heart beat data sample comprising one heart beat cycle;
a second extraction module 402, configured to perform feature extraction on the heartbeat set by using a heartbeat feature extraction model, so as to obtain heartbeat features of the ventricular fibrillation data sample; the heart beat feature extraction model is obtained by utilizing arrhythmia samples for pre-training;
a training module 403, configured to train the ventricular fibrillation detection model based on the heart beat features of the ventricular fibrillation data samples and the heart rhythm features extracted in advance, to obtain a trained ventricular fibrillation detection model.
In an alternative embodiment, the apparatus further comprises:
the third extraction module is used for extracting HRV (high resolution video) features of the ventricular fibrillation data samples; wherein the HRV characteristic belongs to the heart rhythm characteristic.
In addition, the apparatus further comprises:
the fourth extraction module is used for extracting IMF characteristics of the ventricular fibrillation data samples; wherein the IMF characteristic belongs to the heart rhythm characteristic.
Corresponding to the foregoing embodiment, an embodiment of the present application further provides a training method for a ventricular fibrillation detection model, where the ventricular fibrillation detection model is used in the ventricular fibrillation detection apparatus, and referring to fig. 5, a flowchart of the training method for the ventricular fibrillation detection model provided in the embodiment of the present application is shown, where the method includes:
s501: extracting a preset number of heartbeat data from ventricular fibrillation data samples to form a heartbeat set of the ventricular fibrillation data samples; wherein the ventricular fibrillation data samples comprise a plurality of heart beat cycles, each heart beat data sample comprising one heart beat cycle;
s502: performing feature extraction on the heart beat set by using a heart beat feature extraction model to obtain heart beat features of the ventricular fibrillation data sample; the heart beat feature extraction model is obtained by utilizing arrhythmia samples to pre-train;
s503: and training the ventricular fibrillation detection model based on the heart beat features of the ventricular fibrillation data samples and the heart rhythm features extracted in advance to obtain the trained ventricular fibrillation detection model.
In an optional embodiment, the method further comprises:
extracting HRV characteristics of the ventricular fibrillation data sample; wherein the HRV characteristic belongs to the heart rhythm characteristic.
In another optional embodiment, the method further comprises:
extracting IMF characteristics of the ventricular fibrillation data sample; wherein the IMF characteristic belongs to the heart rhythm characteristic.
According to the training method and device for the ventricular fibrillation detection model, the cardiac beat feature extraction model is pre-trained by using a large number of arrhythmia data samples, so that the training precision of the cardiac beat feature extraction model can be guaranteed under the condition that the ventricular fibrillation data samples are insufficient. The ventricular fibrillation detection model is trained based on heart beat features and heart rhythm features extracted by the heart beat feature extraction model, so that a more accurate ventricular fibrillation detection model can be obtained, and the accuracy of ventricular fibrillation detection is improved.
In addition, the present application further provides a ventricular fibrillation detection apparatus, as shown in fig. 6, which may include:
a processor 601, a memory 602, an input device 603, and an output device 604. The number of processors 601 in the ventricular fibrillation detection apparatus may be one or more, and one processor is taken as an example in fig. 6. In some embodiments of the invention, the processor 601, the memory 602, the input device 603 and the output device 604 may be connected by a bus or other means, wherein the connection by the bus is exemplified in fig. 6.
The memory 602 may be used to store software programs and modules, and the processor 601 may execute various functional applications and data processing of the ventricular fibrillation detection device by executing the software programs and modules stored in the memory 602. The memory 602 may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application program required for at least one function, and the like. Further, the memory 602 may include high speed random access memory, and may also include non-volatile memory, such as at least one magnetic disk storage device, flash memory device, or other volatile solid state storage device. The input device 603 may be used to receive entered numerical or character information and to generate signal inputs related to user settings and functional control of the ventricular fibrillation detection apparatus.
Specifically, in this embodiment, the processor 601 loads an executable file corresponding to a process of one or more application programs into the memory 602 according to the following instructions, and the processor 601 runs the application programs stored in the memory 602, thereby implementing various functions of the ventricular fibrillation detection apparatus.
In addition, the present application also provides a computer-readable storage medium having stored therein instructions that, when executed on a terminal device, cause the terminal device to implement ventricular fibrillation detection functionality.
In addition, an embodiment of the present application further provides a training device for a ventricular fibrillation detection model, as shown in fig. 7, the training device may include:
a processor 701, a memory 702, an input device 703, and an output device 704. The number of processors 701 in the training apparatus for the ventricular fibrillation detection model may be one or more, and one processor is taken as an example in fig. 7. In some embodiments of the invention, the processor 701, the memory 702, the input device 703 and the output device 704 may be connected by a bus or other means, wherein the connection by the bus is exemplified in fig. 7.
The memory 702 may be used to store software programs and modules, and the processor 701 may perform various functional applications and data processing of the training apparatus for the ventricular fibrillation detection model by operating the software programs and modules stored in the memory 702. The memory 702 may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application program required for at least one function, and the like. Further, the memory 702 may include high speed random access memory, and may also include non-volatile memory, such as at least one magnetic disk storage device, flash memory device, or other volatile solid state storage device. The input device 703 may be used to receive input numeric or character information and to generate signal inputs related to user settings and function controls of the training apparatus for the ventricular fibrillation detection model.
Specifically, in this embodiment, the processor 701 may load an executable file corresponding to a process of one or more application programs into the memory 702 according to the following instructions, and the processor 701 runs the application programs stored in the memory 702, so as to implement various functions of the training apparatus of the ventricular fibrillation detection model.
In addition, the application also provides a computer readable storage medium, which stores instructions that, when executed on a terminal device, enable the terminal device to implement a training function of a ventricular fibrillation detection model.
It is understood that for the apparatus embodiments, since they correspond substantially to the method embodiments, reference may be made to the partial description of the method embodiments for relevant points. The above-described embodiments of the apparatus are merely illustrative, and the units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of the present embodiment. One of ordinary skill in the art can understand and implement without inventive effort.
It should be noted that, in this document, relational terms such as first and second, and the like are used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Also, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrases "comprising one of 8230; \8230;" 8230; "does not exclude the presence of additional like elements in a process, method, article, or apparatus that comprises the element.
The above detailed descriptions of the detecting device for ventricular fibrillation, the training device for a ventricular fibrillation detecting model, the method and the related apparatus provided by the embodiments of the present application have been provided, and specific examples are applied herein to illustrate the principles and embodiments of the present application, and the above descriptions of the embodiments are only used to help understanding the method and the core ideas of the present application; meanwhile, for a person skilled in the art, according to the idea of the present application, there may be variations in the specific embodiments and the application scope, and in summary, the content of the present specification should not be construed as a limitation to the present application.
Claims (14)
1. A ventricular fibrillation detection device, characterized in that it comprises:
the first extraction unit is used for extracting a preset number of heartbeat data from the heart rate data to be detected to form a heartbeat set of the heart rate data to be detected; the heart rate data to be detected comprises a plurality of heart beat cycles, and each heart beat data comprises one heart beat cycle;
the second extraction unit is used for extracting the features of the heart beat set by using a heart beat feature extraction model to obtain heart beat features of the heart beat data to be detected; the heart beat feature extraction model is obtained by training a pre-training model by using ventricular fibrillation data samples, the pre-training model is obtained by training arrhythmia data samples, and the ventricular fibrillation data samples comprise normal heart beat data samples and ventricular fibrillation data samples;
and the detection unit is used for inputting the heart beat characteristics and the heart rhythm characteristics extracted in advance into a trained ventricular fibrillation detection model to obtain a detection result of the heart rhythm data to be detected, wherein the trained ventricular fibrillation detection model is obtained by training the ventricular fibrillation detection model by utilizing the heart beat characteristics and the heart rhythm characteristics of a ventricular fibrillation data sample extracted by the heart beat characteristic extraction model.
2. The apparatus of claim 1, further comprising:
the third extraction unit is used for extracting the HRV (heart rate variability) characteristics of the heart rate data to be detected; wherein the HRV characteristic belongs to the heart rhythm characteristic.
3. The apparatus of claim 1 or 2, further comprising:
a fourth extraction unit, configured to extract an IMF feature of the cardiac rhythm data to be detected; wherein the IMF characteristic belongs to the heart rhythm characteristic.
4. The apparatus of claim 3, wherein the fourth extraction unit comprises:
the first extraction subunit is used for extracting the time domain IMF characteristics of the heart rhythm data to be detected;
and/or the second extraction subunit is used for extracting the frequency domain IMF characteristics of the heart rhythm data to be detected.
5. The apparatus of claim 4, wherein the first extraction subunit comprises:
the third extraction subunit is used for extracting the IMF1 signal and the residual signal of the heart rate data to be detected;
and the first calculating subunit is used for calculating the similarity between the IMF1 signal and the residual signal and the to-be-detected rhythm data respectively in a time domain range, and the similarity is used as the time domain IMF characteristics of the to-be-detected rhythm data.
6. The apparatus of claim 4, wherein the second extraction subunit comprises:
the fourth extraction subunit is used for extracting the IMF1 signal and the residual signal of the heart rhythm data to be detected;
and the second calculating subunit is used for calculating the similarity between the IMF1 signal and the residual signal and the data of the heart rhythm to be detected respectively in a frequency domain range, and the similarity is used as the frequency domain IMF characteristic of the data of the heart rhythm to be detected.
7. Training device for ventricular fibrillation detection models, characterized in that said ventricular fibrillation detection models are used in a ventricular fibrillation detection device according to any one of claims 1-6 and comprise:
the first extraction module is used for extracting a preset number of heartbeat data from the ventricular fibrillation data samples to form a heartbeat set of the ventricular fibrillation data samples; wherein the ventricular fibrillation data samples comprise a plurality of heart beat cycles, each heart beat data comprises one heart beat cycle, and the ventricular fibrillation data samples comprise normal heart beat data samples and ventricular fibrillation data samples;
the second extraction module is used for performing feature extraction on the heartbeat set by using a heartbeat feature extraction model to obtain heartbeat features of the ventricular fibrillation data samples; the heart beat feature extraction model is obtained by utilizing arrhythmia samples for pre-training;
and the training module is used for training the ventricular fibrillation detection model based on the heart beat characteristics of the ventricular fibrillation data samples and the heart rhythm characteristics extracted in advance to obtain the trained ventricular fibrillation detection model.
8. The apparatus of claim 7, further comprising:
the third extraction module is used for extracting HRV (high resolution video) features of the ventricular fibrillation data samples; wherein the HRV characteristic belongs to the heart rhythm characteristic.
9. The apparatus of claim 7 or 8, further comprising:
the fourth extraction module is used for extracting IMF characteristics of the ventricular fibrillation data samples; wherein the IMF characteristic belongs to the heart rhythm characteristic.
10. A method for training a ventricular fibrillation detection model, wherein the ventricular fibrillation detection model is used in a ventricular fibrillation detection device according to any one of claims 1-6, and the method comprises the following steps:
extracting a preset number of heartbeat data from ventricular fibrillation data samples to form a heartbeat set of the ventricular fibrillation data samples; wherein the ventricular fibrillation data samples comprise a plurality of heart beat cycles, each heart beat data sample comprising a heart beat cycle, the ventricular fibrillation data samples comprise normal heart beat data samples and ventricular fibrillation data samples;
performing feature extraction on the heart beat set by using a heart beat feature extraction model to obtain heart beat features of the ventricular fibrillation data sample; the heart beat feature extraction model is obtained by utilizing arrhythmia samples for pre-training;
and training the ventricular fibrillation detection model based on the heart beat features of the ventricular fibrillation data samples and the heart rhythm features extracted in advance to obtain the trained ventricular fibrillation detection model.
11. The method of claim 10, further comprising: extracting HRV characteristics of the ventricular fibrillation data sample; wherein the HRV characteristic belongs to the heart rhythm characteristic.
12. The method according to claim 10 or 11, characterized in that the method further comprises:
extracting IMF characteristics of the ventricular fibrillation data sample; wherein the IMF characteristic belongs to the heart rhythm characteristic.
13. A computer-readable storage medium having stored therein instructions which, when run on a terminal device, cause the terminal device to implement the functionality of the apparatus of any one of claims 1-9 or to perform the method of any one of claims 10-12.
14. An apparatus, comprising: memory, a processor, and a computer program stored on the memory and executable on the processor, the processor implementing the functionality of the apparatus of any of claims 1-9 or performing the method of any of claims 10-12 when executing the computer program.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911237611.7A CN111067512B (en) | 2019-12-05 | 2019-12-05 | Ventricular fibrillation detection device, ventricular fibrillation detection model training method and equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911237611.7A CN111067512B (en) | 2019-12-05 | 2019-12-05 | Ventricular fibrillation detection device, ventricular fibrillation detection model training method and equipment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111067512A CN111067512A (en) | 2020-04-28 |
| CN111067512B true CN111067512B (en) | 2023-01-31 |
Family
ID=70313272
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911237611.7A Active CN111067512B (en) | 2019-12-05 | 2019-12-05 | Ventricular fibrillation detection device, ventricular fibrillation detection model training method and equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111067512B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113349753A (en) * | 2021-07-19 | 2021-09-07 | 成都芯跳医疗科技有限责任公司 | Arrhythmia detection method based on portable dynamic electrocardiogram monitor |
| CN114334090B (en) * | 2022-03-02 | 2022-07-12 | 博奥生物集团有限公司 | Data analysis method and device and electronic equipment |
| CN114869249B (en) * | 2022-07-12 | 2022-11-01 | 慧医谷中医药科技(天津)股份有限公司 | Method and device for acquiring pulse wave characteristic parameters, electronic equipment and medium |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104881631A (en) * | 2015-04-16 | 2015-09-02 | 广西师范大学 | Multi-characteristic integrated cervical cell image characteristic extraction and identification method, and cervical cell characteristic identification device |
| CN108309262A (en) * | 2018-02-24 | 2018-07-24 | 乐普（北京）医疗器械股份有限公司 | Multi-parameter monitoring data analysing method and multi-parameter monitor |
| CN109171712A (en) * | 2018-09-28 | 2019-01-11 | 东软集团股份有限公司 | Auricular fibrillation recognition methods, device, equipment and computer readable storage medium |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9451898B2 (en) * | 2014-07-02 | 2016-09-27 | National Central University | Method and system for extracting ventricular fibrillation signals in electrocardiogram using spline interpolation with uniform phase ensembles |
-
2019
- 2019-12-05 CN CN201911237611.7A patent/CN111067512B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104881631A (en) * | 2015-04-16 | 2015-09-02 | 广西师范大学 | Multi-characteristic integrated cervical cell image characteristic extraction and identification method, and cervical cell characteristic identification device |
| CN108309262A (en) * | 2018-02-24 | 2018-07-24 | 乐普（北京）医疗器械股份有限公司 | Multi-parameter monitoring data analysing method and multi-parameter monitor |
| CN109171712A (en) * | 2018-09-28 | 2019-01-11 | 东软集团股份有限公司 | Auricular fibrillation recognition methods, device, equipment and computer readable storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111067512A (en) | 2020-04-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Gao et al. | An effective LSTM recurrent network to detect arrhythmia on imbalanced ECG dataset | |
| KR102451795B1 (en) | ECG signal detection method | |
| Jin et al. | Multi-domain modeling of atrial fibrillation detection with twin attentional convolutional long short-term memory neural networks | |
| Ahmad et al. | ECG heartbeat classification using multimodal fusion | |
| CN111067512B (en) | Ventricular fibrillation detection device, ventricular fibrillation detection model training method and equipment | |
| CN111160139B (en) | Electrocardiosignal processing method and device and terminal equipment | |
| Wang et al. | Arrhythmia classification algorithm based on multi-head self-attention mechanism | |
| Deng et al. | Extracting cardiac dynamics within ECG signal for human identification and cardiovascular diseases classification | |
| CN107890348B (en) | One kind is extracted based on the automation of deep approach of learning electrocardio tempo characteristic and classification method | |
| WO2021017313A1 (en) | Atrial fibrillation detection method and apparatus, computer device, and storage medium | |
| Luo et al. | Multi-classification of arrhythmias using a HCRNet on imbalanced ECG datasets | |
| CN110619322A (en) | Multi-lead electrocardio abnormal signal identification method and system based on multi-flow convolution cyclic neural network | |
| Singh et al. | Classification of short unsegmented heart sound based on deep learning | |
| KR20070009667A (en) | Cardiac diagnostic system and method | |
| Wang et al. | Automatic classification of CAD ECG signals with SDAE and bidirectional long short-term network | |
| CN111488793A (en) | Electrocardiosignal classification method and device, electronic equipment and storage medium | |
| US20180177415A1 (en) | Cardiovascular disease detection | |
| Sanjana et al. | Explainable artificial intelligence for heart rate variability in ECG signal | |
| WO2022166158A1 (en) | System for performing long-term hazard prediction on hemodialysis complications on basis of convolutional survival network | |
| Chen et al. | Heart sound segmentation via duration long–short term memory neural network | |
| Li et al. | Signal processing and feature selection preprocessing for classification in noisy healthcare data | |
| CN112426160A (en) | Electrocardiosignal type identification method and device | |
| Rohmantri et al. | Arrhythmia classification using 2D convolutional neural network | |
| Chen et al. | Edge2Analysis: a novel AIoT platform for atrial fibrillation recognition and detection | |
| CN109124620A (en) | A kind of atrial fibrillation detection method, device and equipment |
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 |