CN109480827B - Vector electrocardiogram classification method and device - Google Patents
Vector electrocardiogram classification method and device Download PDFInfo
- Publication number
- CN109480827B CN109480827B CN201811551924.5A CN201811551924A CN109480827B CN 109480827 B CN109480827 B CN 109480827B CN 201811551924 A CN201811551924 A CN 201811551924A CN 109480827 B CN109480827 B CN 109480827B
- Authority
- CN
- China
- Prior art keywords
- electrocardiogram
- neural network
- dimensional
- network model
- vector
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
- A61B5/316—Modalities, i.e. specific diagnostic methods
- A61B5/318—Heart-related electrical modalities, e.g. electrocardiography [ECG]
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7235—Details of waveform analysis
- A61B5/7264—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems
- A61B5/7267—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems involving training the classification device
Landscapes
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Physics & Mathematics (AREA)
- Biophysics (AREA)
- Surgery (AREA)
- Veterinary Medicine (AREA)
- Public Health (AREA)
- General Health & Medical Sciences (AREA)
- Animal Behavior & Ethology (AREA)
- Molecular Biology (AREA)
- Medical Informatics (AREA)
- Pathology (AREA)
- Biomedical Technology (AREA)
- Heart & Thoracic Surgery (AREA)
- Evolutionary Computation (AREA)
- Fuzzy Systems (AREA)
- Mathematical Physics (AREA)
- Signal Processing (AREA)
- Psychiatry (AREA)
- Physiology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Cardiology (AREA)
- Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)
Abstract
The invention provides a vector electrocardiogram classification method and a vector electrocardiogram classification device, which relate to the technical field of vector electrocardiogram classification, and can be used for respectively projecting a three-dimensional electrocardiogram to a two-dimensional plane of a current lead system when a three-dimensional electrocardiogram to be classified is obtained to generate an electrocardiogram vector diagram corresponding to the three-dimensional electrocardiogram; extracting a three-dimensional electrocardiogram ring contained in the three-dimensional electrocardiogram to be classified; respectively inputting the vector electrocardiogram and the three-dimensional electrocardiogram into a first neural network model and a second neural network model, and extracting potential direction and amplitude change contained in the three-dimensional electrocardiogram; inputting the output results of the first neural network model and the second neural network model into a third neural network model for feature fusion; the stereo electrocardiogram is classified according to the fusion result, and the technical problem of wrong diagnosis conclusion caused by unclear waveform characteristic points in the traditional diagnosis electrocardiogram is effectively solved.
Description
Technical Field
The invention relates to the technical field of vector electrocardiogram classification, in particular to a vector electrocardiogram classification method and device.
Background
The heart is a three-dimensional organ, when the cardiac electrode is removed, the upper part, the lower part, the left part, the right part, the front part and the back part of the heart are almost simultaneously removed, and the three-dimensional electrocardiogram shows the change of the cardiac electrode removal vector and the repolarization vector at a certain moment. The stereo electrocardiogram can comprehensively and finely reflect the depolarization direction and sequence of the heart and the change of a stereo space. One plane cannot reflect all the potential changes of the three planes, and only the projection of the three planes can form a stereo, dynamic and full-coverage heart potential change process. The electrocardiogram is formed by projecting stereo electrocardiogram to transverse plane, frontal plane and lateral plane to form corresponding electrocardiogram vector loop, and then performing secondary projection to form corresponding twelve-lead electrocardiogram time sequence waveform. The electrocardiogram is the electrophysiological activity of cardiac muscle formed by conducting a series of potential changes to various parts of the heart through a conduction system after the sinus node performs automatic pacing. According to the time sequence of the activation of the heart, the change of the body surface potential is recorded, and the formed continuous curve is the electrocardiogram. Typical electrocardiograms include P-wave, QRS-wave, T-wave. The P wave reflects the potential change in the atrial depolarization process; the P-R interval represents the time period from activation of the sinoatrial node through the atrioventricular junction to the onset of depolarization of the ventricular muscle; the QRS complex reflects the potential change in the ventricular depolarization process; the T wave represents the potential change during repolarization of the ventricular muscle.
In the traditional electrocardiogram diagnosis, common characteristic parameters are parameters such as RR interval, PR interval, QT interval, P, QRS and T wave electric axes obtained by positioning the starting position and the ending position of P, QRS and T wave, when the waveform morphological characteristics are obvious, the detection effect is good, and when the characteristic points of the waveform are not clear, the parameters are easy to detect errors, so that the error of a diagnosis conclusion can be caused.
Disclosure of Invention
In view of the above, the present invention provides a method and an apparatus for classifying a vector electrocardiogram, so as to alleviate the technical problem that the waveform feature points of the conventional diagnostic electrocardiogram are ambiguous and thus easily cause errors in diagnostic results.
In a first aspect, an embodiment of the present invention provides a vector electrocardiogram classification method, where the method includes: acquiring a three-dimensional electrocardiogram to be classified, and projecting the three-dimensional electrocardiogram onto a two-dimensional plane of a current lead system respectively to generate an electrocardiogram vector diagram corresponding to the three-dimensional electrocardiogram; extracting a three-dimensional electrocardiogram ring contained in the three-dimensional electrocardiogram to be classified; respectively inputting the vector electrocardiogram and the three-dimensional electrocardiogram into a first neural network model and a second neural network model, and extracting potential direction and amplitude change contained in the three-dimensional electrocardiogram; inputting the output results of the first neural network model and the second neural network model into a third neural network model for feature fusion; classifying the stereocardiograms according to the fusion result; the first neural network model, the second neural network model and the third neural network model are electrocardio data classification models obtained by training electrocardiogram samples carrying diagnostic markers.
With reference to the first aspect, an embodiment of the present invention provides a first possible implementation manner of the first aspect, where the stereo electrocardiogram is an electrocardiographic signal acquired under a Frank lead system; the step of obtaining a stereoscopic electrocardiogram to be classified comprises: and acquiring potential change on coordinate axes included by the Frank lead system to generate a three-dimensional electrocardiogram, wherein the coordinate axes include an X axis, a Y axis and a Z axis which are mutually perpendicular.
With reference to the first possible implementation manner of the first aspect, an embodiment of the present invention provides a second possible implementation manner of the first aspect, where the vector cardiac diagram is generated by projecting a stereo electrocardiogram onto two-dimensional planes XY, YZ, and XZ, respectively.
With reference to the second possible implementation manner of the first aspect, an embodiment of the present invention provides a third possible implementation manner of the first aspect, where the first neural network model includes a plurality of convolutional networks and a first fully-connected layer; the convolution networks comprise a first convolution network, a second convolution network and a third convolution network which extract an electrocardiogram projected on two-dimensional planes of XY, YZ and XZ respectively; the fourth convolution network, the fifth convolution network and the sixth convolution network are used for fusing output results of the first convolution network, the second convolution network and the third convolution network; the output results of the first convolution network and the second convolution network are input into a fourth convolution network; the output results of the first convolution network and the third convolution network are input into a fifth convolution network; the output results of the second convolution network and the third convolution network are input into a sixth convolution network; and the output results of the fourth convolution network, the fifth convolution network and the sixth convolution network are input to the first full connection layer.
With reference to the first aspect, an embodiment of the present invention provides a fourth possible implementation manner of the first aspect, where the second neural network model includes a three-dimensional convolutional network and a second fully-connected layer connected to the three-dimensional convolutional network.
With reference to the first aspect, an embodiment of the present invention provides a fifth possible implementation manner of the first aspect, where the third neural network model includes a third fully-connected layer, a fourth fully-connected layer, and an output classification layer, which are connected in sequence.
In a second aspect, an embodiment of the present invention further provides a vector electrocardiogram classification apparatus, where the apparatus includes: the acquisition module is used for acquiring the stereoscopic electrocardiograms to be classified, and projecting the stereoscopic electrocardiograms onto a two-dimensional plane of a current lead system respectively to generate an electrocardiovector diagram corresponding to the stereoscopic electrocardiograms; the first extraction module is used for extracting the three-dimensional electrocardiogram loops contained in the three-dimensional electrocardiogram to be classified; the second extraction module is used for respectively inputting the vector cardiogram and the three-dimensional electrocardiogram into the first neural network model and the second neural network model and extracting the potential direction and amplitude change contained in the three-dimensional electrocardiogram; the fusion module is used for inputting the output results of the first neural network model and the second neural network model into the third neural network model for feature fusion; the classification module is used for classifying the stereocardiogram according to the fusion result; the first neural network model, the second neural network model and the third neural network model are electrocardio data classification models obtained by training electrocardiogram samples carrying diagnostic markers.
With reference to the second aspect, the embodiments of the present invention provide a first possible implementation manner of the second aspect, wherein the stereo electrocardiogram is an electrocardiographic signal acquired under a Frank lead system; the acquisition module is further configured to: and acquiring potential change on coordinate axes included by the Frank lead system to generate a three-dimensional electrocardiogram, wherein the coordinate axes include an X axis, a Y axis and a Z axis which are mutually perpendicular.
With reference to the first possible implementation manner of the second aspect, the embodiment of the present invention provides a second possible implementation manner of the second aspect, wherein the vector cardiac diagram is generated by projecting a stereo electrocardiogram onto two-dimensional planes XY, YZ, and XZ, respectively.
In a third aspect, an embodiment of the present invention further provides a computer storage medium, configured to store computer program instructions, where the computer program instructions, when executed by a computer, perform the above-mentioned method.
The embodiment of the invention has the following beneficial effects:
the vector electrocardiogram classification method and device provided by the embodiment of the invention can respectively project the three-dimensional electrocardiograms onto the two-dimensional plane of the current lead system when the three-dimensional electrocardiograms to be classified are obtained, so as to generate the corresponding vector electrocardiograms of the three-dimensional electrocardiograms; extracting a three-dimensional electrocardiogram ring contained in the three-dimensional electrocardiogram to be classified; respectively inputting the vector electrocardiogram and the three-dimensional electrocardiogram into a first neural network model and a second neural network model, and extracting potential direction and amplitude change contained in the three-dimensional electrocardiogram; inputting the output results of the first neural network model and the second neural network model into a third neural network model for feature fusion; the stereo electrocardiogram is classified according to the fusion result, and the technical problem of wrong diagnosis conclusion caused by unclear waveform characteristic points in the traditional diagnosis electrocardiogram is effectively solved.
Additional features and advantages of the invention will be set forth in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. The objectives and other advantages of the invention will be realized and attained by the structure particularly pointed out in the written description and drawings.
In order to make the aforementioned and other objects, features and advantages of the present invention comprehensible, preferred embodiments accompanied with figures are described in detail below.
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, and it is obvious that the drawings in the following description are some embodiments of the present invention, and other drawings can be obtained by those skilled in the art without creative efforts.
Fig. 1 is a flowchart of a vector electrocardiogram classification method according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of a time vector cardiogram according to an embodiment of the present invention;
FIG. 3 is a diagram illustrating feature extraction using a first neural network model according to an embodiment of the present invention;
FIG. 4 is a schematic diagram of feature extraction using a second neural network model according to an embodiment of the present invention;
FIG. 5 is a diagram illustrating classification using a third neural network model according to an embodiment of the present invention;
fig. 6 is a schematic structural diagram of a vector electrocardiogram classification apparatus according to an embodiment of the present invention.
Detailed Description
To make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings, and it is apparent that the described embodiments are some, but not all embodiments of the present invention. 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.
At present, medical experts determine a diagnosis result through electrocardiogram analysis, and because the characteristic points of an electrocardiogram waveform are not clear, the technical problem of errors of a diagnosis conclusion is easily caused.
To facilitate understanding of the present embodiment, a method for classifying a vector electrocardiogram disclosed in the present embodiment will be described in detail.
The first embodiment is as follows:
the embodiment of the invention provides a vector electrocardiogram classification method, such as a flow chart of the vector electrocardiogram classification method shown in figure 1, which comprises the following steps:
step S102, obtaining a three-dimensional electrocardiogram to be classified, projecting the three-dimensional electrocardiogram onto a two-dimensional plane of a current lead system respectively, and generating an electrocardiogram corresponding to the three-dimensional electrocardiogram;
step S104, extracting three-dimensional electrocardio-rings contained in the three-dimensional electrocardiograms to be classified;
generally, the stereoscopic electrocardiogram comprehensively reflects the depolarization process of the heart and the stereoscopic change process of the electrocardiogram, contains rich electrocardiogram potential information, and when myocardial cells are pathologically changed, the potentials at corresponding positions can also change. The vector cardiogram records the heart electrical activation process of one cardiac cycle. The temporal relationship between the atrioventricular chambers and between each heart rhythm cannot be displayed. And thus certain arrhythmias may not be discernable. The time vector cardiogram can make up for the defects. The overlapped P ring, QRS ring and T ring of the vector cardiogram are unfolded along the time axis, a plurality of cardiac cycles can be recorded, the time of each ring, the time relation among the rings and the time relation among the cardiac cycles can be measured, and the recorded three rings of the projections of a plurality of continuous different cycles on the frontal plane, the transverse plane and the right side form a 'time vector cardiogram'. The time vector cardiogram integrates the two-dimensional curve of the vector cardiogram and the function of continuously recording a plurality of cardiac cycles by the electrocardiogram, can diagnose arrhythmia, and displays the average QRS axis vector and the initial vector of the QRS ring more clearly than the common vector. The stereo electrocardiograph ring contains rich information of potential change of the heart without losing information which is not contained in a 12-lead electrocardiogram and a time vector electrocardiogram. The three-dimensional electrocardiogram adopts a Frank lead system, the lead system has 7 electrodes, 5 electrodes are placed on the chest, the back of the neck is deviated to the right 1cm, the left foot and the 7 leads form three X, Y, Z axes which are vertical to each other, the X axis and the Y axis form a forehead plane F, the Y axis and the Z axis form a right side surface S, the X axis and the Z axis form a transverse plane H, the Frank lead can be used for simultaneously obtaining X, Y, Z lead electrocardiosignals on three axes, and the lead electrocardiosignals on the three axes can be respectively obtained by the following calculation:
Vx＝0.610×V4+0.171×V3-0.781×V1；
Vy＝0.345×V5+0.655×F-V6；
Vz＝0.133×V4+0.736×V5-0.264×V1-0.374×V2-0.231×V3；
wherein, Vx represents lead electrocardiosignal of X axis, Vy represents lead electrocardiosignal of Y axis, Vz represents lead electrocardiosignal of Z axis, V1 to V5 represent potential signal of 5 electrodes placed on chest respectively, V6 represents potential signal of electrode placed on the position 1cm away from the right of the back of neck, F represents potential signal of electrode placed on the left foot; the sequential electrocardiosignals of X, Y, Z axes obtained by a Frank lead system are continuously represented in a three-dimensional space to obtain a stereogram, and the stereogram is projected on two-dimensional planes of XY, YZ and XZ respectively to obtain an electrocardiogram, namely the process of decomposing the stereogram of the electrocardiosignals of the electrical excitation, namely weak current, generated by the heart of a human body into three planes projected in a X, Y, Z three-dimensional space coordinate system is realized. Therefore, the frontal modulus of the frontal plane time vector cardiogram obtained by projecting the stereo vector cardiogram onto the XY two-dimensional plane is as follows:
the azimuth angle is:
the side modulus of a side time vector cardiogram obtained by projecting a stereo vector electrocardiogram onto a YZ two-dimensional plane is as follows:
the azimuth angle is:
the transverse modulus of a transverse time vector cardiogram obtained by projecting the stereo vector cardiogram to an XZ two-dimensional plane is as follows:
the azimuth angle is:
wherein t is the acquisition time in units of s. And the time-sequenced X, Y, Z-axis lead electrocardiosignals constitute a stereo vector electrocardio-ring.
Step S106, inputting the vector cardiogram and the three-dimensional electrocardiogram into a first neural network model and a second neural network model respectively, and extracting potential direction and amplitude variation contained in the three-dimensional electrocardiogram;
step S108, inputting output results of the first neural network model and the second neural network model into a third neural network model for feature fusion;
step S110, classifying the stereocardiograms according to the fusion result; the first neural network model, the second neural network model and the third neural network model are electrocardio data classification models obtained by training electrocardiogram samples carrying diagnostic markers.
In the concrete implementation, 3 sequences of time vector cardiograms are selected on XY, YZ and XZ two-dimensional planes according to preset sampling rate and sampling time respectively, in order to facilitate understanding of the 3 sequences of time vector cardiograms, FIG. 2 shows a schematic diagram of a time vector cardiogram, for example, FIG. 2 shows a frontal time vector cardiogram, a side time vector cardiogram and a transverse time vector cardiogram respectively, the characteristics of potential direction and amplitude change contained in the 3 time vector cardiograms are extracted by using a first neural network model respectively, then the characteristics of potential direction and amplitude change contained in a three-dimensional electrocardiogram are extracted by using a second neural network model, and then the characteristics extracted by using a third neural network model and the characteristics extracted by the second neural network model are subjected to fusion analysis, and finally, obtaining the classification result of the to-be-detected three-dimensional electrocardiogram.
The vector electrocardiogram classification method provided by the embodiment of the invention can be used for respectively projecting the three-dimensional electrocardiograms onto a two-dimensional plane of a current lead system when the three-dimensional electrocardiograms to be classified are obtained, so as to generate the vector electrocardiograms corresponding to the three-dimensional electrocardiograms; extracting a three-dimensional electrocardiogram ring contained in the three-dimensional electrocardiogram to be classified; respectively inputting the vector electrocardiogram and the three-dimensional electrocardiogram into a first neural network model and a second neural network model, and extracting potential direction and amplitude change contained in the three-dimensional electrocardiogram; inputting the output results of the first neural network model and the second neural network model into a third neural network model for feature fusion; the stereo electrocardiogram is classified according to the fusion result, and the technical problem of wrong diagnosis conclusion caused by unclear waveform characteristic points in the traditional diagnosis electrocardiogram is effectively solved.
Further, the first neural network model includes a plurality of convolutional networks and a first fully-connected layer; the convolution networks comprise a first convolution network, a second convolution network and a third convolution network which extract an electrocardiogram projected on two-dimensional planes of XY, YZ and XZ respectively; the fourth convolution network, the fifth convolution network and the sixth convolution network are used for fusing output results of the first convolution network, the second convolution network and the third convolution network; the output results of the first convolution network and the second convolution network are input into a fourth convolution network; the output results of the first convolution network and the third convolution network are input into a fifth convolution network; the output results of the second convolution network and the third convolution network are input into a sixth convolution network; and the output results of the fourth convolution network, the fifth convolution network and the sixth convolution network are input to the first full connection layer.
To facilitate understanding of the structure of the first Neural network model, fig. 3 shows a schematic diagram of extracting features using the first Neural network model, and as shown in fig. 3, CNN1(Convolutional Neural Networks) to CNN5 represent the first Convolutional network to the fifth Convolutional network, respectively. Preferably, the structural parameters of each convolutional neural network of CNN1, CNN2, and CNN3 are set as:
(1) the first layer of convolution: the size of the convolution kernel is 1 multiplied by 3, the step length is set to be 2, the number of the convolution kernels is 16, the output is 1 multiplied by 2498 multiplied by 16, and the activation function is relu;
(2) first-layer pooling: the average pooling of 1 × 2 is adopted, and the output is 1 × 1249 × 16;
(3) second layer convolution: the size of the convolution kernel is 1 multiplied by 3, the step length is set to be 2, the number of the convolution kernels is 32, the output is 1 multiplied by 622 multiplied by 32, and the activation function is relu;
(4) second-layer pooling: the average pooling of 1 × 2 is adopted, and the output is 1 × 311 × 32;
(5) and a third layer of convolution: the size of the convolution kernel is 1 multiplied by 3, the step length is set to be 2, the number of the convolution kernels is 64, the output is 1 multiplied by 154 multiplied by 64, and the activation function is relu;
(6) and (3) third-layer pooling: the average pooling of 1 × 2 is adopted, and the output is 1 × 77 × 64;
(7) and a fourth layer of convolution: the size of the convolution kernel is 1 multiplied by 3, the step length is set to be 2, the number of the convolution kernels is 128, the output is 1 multiplied by 36 multiplied by 128, and the activation function is relu;
(8) and fourth-layer pooling: the output was 1 × 19 × 128 with an average pooling of 1 × 2.
The structural parameters of each convolutional neural network of CNN4, CNN5, and CNN6 are set as:
(1) the first layer of convolution: the size of the convolution kernel is 2 multiplied by 2, the step length is set to be 1, the number of the convolution kernels is 256, the output is 2 multiplied by 18 multiplied by 256, and the activation function is relu;
(2) first-layer pooling: 2 × 2 average pooling is adopted, and the output is 2 × 9;
(3) second layer convolution: the convolution kernel size is 2 × 2, the step size is set to 1, the number of convolution kernels is 512, the output is 2 × 9 × 512, and the activation function is relu.
The structural parameters of the first full-connection layer are set as follows: the number of convolution kernels is 512 and the activation function is relu.
In concrete implementation, the CNN1, CNN2 and CNN3 respectively perform feature extraction on the frontal time vector cardiogram, the lateral time vector cardiogram and the transverse time vector cardiogram, extract features of potential direction and amplitude variation contained in the time vector cardiogram, input output results of CNN1 and CNN2 to CNN4, input output results of CNN1 and CNN3 to CNN5, input output results of CNN2 and CNN3 to CNN6, and further perform feature fusion extraction on different planes; and finally, inputting the feature results extracted by the CNN4, the CNN5 and the CNN6 into the first full connection layer for feature fusion. According to the embodiment of the invention, the structural setting of the first neural network model and the parameter setting of each convolutional neural network are not limited.
Wherein the second neural network model comprises a three-dimensional convolutional network and a second fully connected layer connected with the three-dimensional convolutional network.
In particular implementation, in order to facilitate understanding of the structure of the second neural network model, fig. 4 shows a schematic diagram of feature extraction by using the second neural network model, and as shown in fig. 4, X, Y, Z-axis time series electrocardiographic signals obtained from the Frank lead system, that is, 3 lead electrocardiographic signals of Vx, Vy and Vz of a stereoscopic electrocardiographic loop are input into a three-dimensional convolution network to extract the relationship features between X, Y, Z three axes, and the extracted relationship features are input into the second full-connection layer for feature fusion. The input signals are (Vx, Vy, Vz)10s signals, and the structural parameters of the three-dimensional convolution network are preferably set as follows:
(1) the first layer of convolution: the size of the convolution kernel is 5 multiplied by 5, the step length is set to be 1, the number of the convolution kernels is 16, the output is 1000 multiplied by 1000, and the activation function is relu;
(2) first-layer pooling: the average pooling of 2 × 2 × 2 is adopted, and the output is 500 × 500 × 500;
(3) second layer convolution: the size of the convolution kernel is 5 multiplied by 5, the step length is set to be 1, the number of the convolution kernels is 32, the output is 100 multiplied by 100, and the activation function is relu;
(4) second-layer pooling: the average pooling of 2 × 2 × 2 is adopted, and the output is 50 × 50 × 50;
(5) and a third layer of convolution: the size of the convolution kernel is 5 multiplied by 5, the step length is set to be 1, the number of the convolution kernels is 64, the output is 10 multiplied by 10, and the activation function is relu;
(6) and (3) third-layer pooling: the average pooling of 2X 2 was used and the output was 5X 5.
The structural parameters of the second full-connection layer are set as follows: the number of convolution kernels is 512 and the activation function is relu. According to the embodiment of the invention, the parameter setting of the second neural network model is not limited.
In practical application, the third neural network model comprises a third full connection layer, a fourth full connection layer and an output classification layer which are connected in sequence.
To facilitate understanding of the structure of the third neural network model, fig. 5 shows a schematic diagram of classification using the third neural network model, and as shown in fig. 5, a first fully-connected layer of the first neural network model and a second fully-connected layer of the second neural network model are connected with a third fully-connected layer, which is connected with a fourth fully-connected layer, which is connected with an output classification layer. Preferably, the structural parameters of the third fully-connected layer are set as follows: the number of convolution kernels is 1024, and the activation function is relu; setting the structural parameters of the fourth full connection layer as follows: the number of convolution kernels is 2048, and the activation function is sigmoid; according to the embodiment of the invention, the parameter setting of the third neural network model is not limited.
Example two:
on the basis of the above embodiments, the embodiment of the present invention further provides a vector electrocardiogram classification apparatus, as shown in fig. 6, which includes:
an obtaining module 602, configured to obtain a stereoscopic electrocardiogram to be classified, and respectively project the stereoscopic electrocardiogram onto a two-dimensional plane of a current lead system to generate an vectorcardiogram corresponding to the stereoscopic electrocardiogram; and the number of the first and second groups,
a first extraction module 604, configured to extract a stereo electrocardiograph ring included in a stereo electrocardiogram to be classified;
a second extraction module 606, configured to input the vectorcardiogram and the stereo electrocardiograph to the first neural network model and the second neural network model, respectively, and extract potential direction and amplitude variation included in the stereo electrocardiograph;
a fusion module 608, configured to input output results of the first neural network model and the second neural network model to the third neural network model for feature fusion;
a classification module 610 for classifying the stereocardiogram according to the fused result; the first neural network model, the second neural network model and the third neural network model are electrocardio data classification models obtained by training electrocardiogram samples carrying diagnostic markers.
Wherein, the three-dimensional electrocardiogram is an electrocardiosignal collected under a Frank lead system; the acquisition module is further configured to: and acquiring potential change on coordinate axes included by the Frank lead system to generate a three-dimensional electrocardiogram, wherein the coordinate axes include an X axis, a Y axis and a Z axis which are mutually perpendicular.
Furthermore, the vector cardiogram is generated by projecting the stereogram onto XY, YZ and XZ two-dimensional planes respectively.
The vector electrocardiogram classification device provided by the embodiment of the invention has the same technical characteristics as the vector electrocardiogram classification method provided by the embodiment, so that the same technical problems can be solved, and the same technical effects can be achieved.
Embodiments of the present invention also provide a computer storage medium, which is used for storing computer program instructions, and when the computer executes the computer program instructions, the method described above is performed.
The computer program product of the vector electrocardiogram classification method and apparatus provided in the embodiments of the present invention includes a computer-readable storage medium storing a program code, where instructions included in the program code may be used to execute the method described in the foregoing method embodiments, and specific implementation may refer to the method embodiments, and will not be described herein again.
It is clear to those skilled in the art that, for convenience and brevity of description, the specific working process of the apparatus described above may refer to the corresponding process in the foregoing method embodiment, and is not described herein again.
The functions, if implemented in the form of software functional units and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present invention may be embodied in the form of a software product, which is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and other various media capable of storing program codes.
Finally, it should be noted that: although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art will understand that the following embodiments are merely illustrative of the present invention, and not restrictive, and the scope of the present invention is not limited thereto: any person skilled in the art can modify or easily conceive the technical solutions described in the foregoing embodiments or equivalent substitutes for some technical features within the technical scope of the present disclosure; such modifications, changes or substitutions do not depart from the spirit and scope of the embodiments of the present invention, and they should be construed as being included therein. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.
Claims (10)
1. A method for vector electrocardiogram classification, the method comprising:
acquiring a three-dimensional electrocardiogram to be classified, and respectively projecting the three-dimensional electrocardiogram onto a two-dimensional plane of a current lead system to generate an electrocardiogram vector diagram corresponding to the three-dimensional electrocardiogram; and the number of the first and second groups,
extracting three-dimensional electrocardio-rings contained in the three-dimensional electrocardiogram to be classified;
inputting the vector cardiogram into a first neural network model, extracting potential direction and amplitude variation contained in the vector cardiogram, inputting the stereo electrocardiograph ring into a second neural network model, and extracting potential direction and amplitude variation contained in the stereo electrocardiograph ring;
inputting output results of the first neural network model and the second neural network model into a third neural network model for feature fusion;
classifying the stereoscopic electrocardiogram according to the fused result;
the first neural network model, the second neural network model and the third neural network model are electrocardiogram data classification models obtained by training electrocardiogram samples carrying diagnostic markers.
2. The method according to claim 1, wherein the stereoscopic electrocardiogram is an electrocardiographic signal acquired under the Frank lead system;
the step of obtaining a stereoscopic electrocardiogram to be classified comprises:
and acquiring potential change on coordinate axes included by the Frank lead system to generate a three-dimensional electrocardiogram, wherein the coordinate axes include an X axis, a Y axis and a Z axis which are mutually perpendicular.
3. The method according to claim 2, wherein the vector cardiogram is generated by projecting the stereo electrocardiogram onto XY, YZ, and XZ two-dimensional planes, respectively.
4. The method of claim 3, wherein the first neural network model comprises a plurality of convolutional networks and a first fully-connected layer;
the convolution networks comprise a first convolution network, a second convolution network and a third convolution network which extract the vector cardiograms projected onto two-dimensional planes of XY, YZ and XZ respectively; and a fourth convolution network, a fifth convolution network and a sixth convolution network which fuse the output results of the first convolution network, the second convolution network and the third convolution network;
the output results of the first convolution network and the second convolution network are input into the fourth convolution network;
the output results of the first convolution network and the third convolution network are input into the fifth convolution network;
the output results of the second convolutional network and the third convolutional network are input into the sixth convolutional network;
and the output results of the fourth convolutional network, the fifth convolutional network and the sixth convolutional network are input to the first full-link layer.
5. The method of claim 1, wherein the second neural network model comprises a three-dimensional convolutional network and a second fully-connected layer connected to the three-dimensional convolutional network.
6. The method of claim 1, wherein the third neural network model comprises a third fully-connected layer, a fourth fully-connected layer, and an output classification layer connected in sequence.
7. A vector electrocardiogram classification apparatus, comprising:
the acquisition module is used for acquiring a stereoscopic electrocardiogram to be classified, and projecting the stereoscopic electrocardiogram onto a two-dimensional plane of a current lead system respectively to generate an electrocardiogram vector diagram corresponding to the stereoscopic electrocardiogram; and the number of the first and second groups,
the first extraction module is used for extracting the stereo electrocardiogram contained in the stereo electrocardiogram to be classified;
the second extraction module is used for inputting the vector cardiogram to the first neural network model, extracting potential direction and amplitude variation contained in the vector cardiogram, inputting the stereo electrocardiograph ring to the second neural network model, and extracting potential direction and amplitude variation contained in the stereo electrocardiograph ring;
the fusion module is used for inputting the output results of the first neural network model and the second neural network model into a third neural network model for feature fusion;
the classification module is used for classifying the stereoscopic electrocardiogram according to the fusion result;
the first neural network model, the second neural network model and the third neural network model are electrocardiogram data classification models obtained by training electrocardiogram samples carrying diagnostic markers.
8. The apparatus according to claim 7, wherein the stereoscopic electrocardiogram is an electrocardiographic signal acquired under the Frank lead system;
the acquisition module is further configured to:
and acquiring potential change on coordinate axes included by the Frank lead system to generate a three-dimensional electrocardiogram, wherein the coordinate axes include an X axis, a Y axis and a Z axis which are mutually perpendicular.
9. The apparatus according to claim 8, wherein the vector cardiogram is generated by projecting the electrocardiogram onto XY, YZ, and XZ two-dimensional planes, respectively.
10. A computer storage medium storing computer program instructions for performing the method of any one of claims 1 to 6 when executed by a computer.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811551924.5A CN109480827B (en) | 2018-12-18 | 2018-12-18 | Vector electrocardiogram classification method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811551924.5A CN109480827B (en) | 2018-12-18 | 2018-12-18 | Vector electrocardiogram classification method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109480827A CN109480827A (en) | 2019-03-19 |
| CN109480827B true CN109480827B (en) | 2021-04-16 |
Family
ID=65710849
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811551924.5A Active CN109480827B (en) | 2018-12-18 | 2018-12-18 | Vector electrocardiogram classification method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109480827B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110288028B (en) * | 2019-06-27 | 2021-11-02 | 北京邮电大学 | Electrocardio detection method, system, equipment and computer readable storage medium |
| CN113384277B (en) * | 2020-02-26 | 2022-09-20 | 京东方科技集团股份有限公司 | Electrocardiogram data classification method and classification system |
| CN111710386A (en) * | 2020-04-30 | 2020-09-25 | 上海数创医疗科技有限公司 | Quality control system for electrocardiogram diagnosis report |
| CN111543979B (en) * | 2020-05-13 | 2024-02-06 | 许祥林 | Method for outputting electrocardiographic vector diagram by conventional leads |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN2542204Y (en) * | 2002-05-16 | 2003-04-02 | 南京市玄武区中医医院 | Vector electrocardiograph |
| JP2011143152A (en) * | 2010-01-18 | 2011-07-28 | Kyoto Nishikawa:Kk | Simple negative potential detector |
| CN201912078U (en) * | 2011-01-13 | 2011-08-03 | 深圳市科曼医疗设备有限公司 | Electrocardiogram machine for synchronously outputting electrocardio and electrocardial vector |
| CN102283639A (en) * | 2011-08-31 | 2011-12-21 | 惠州Tcl移动通信有限公司 | Heartbeat detection method, device and system, heartbeat display method and mobile terminal |
| CN104902814A (en) * | 2012-10-12 | 2015-09-09 | 加利福尼亚大学董事会 | Configuration and spatial placement of frontal electrode sensors to detect physiological signals |
| CN107530019A (en) * | 2015-03-31 | 2018-01-02 | 皇家飞利浦有限公司 | Automatic detection/the classification exchanged for the ECG cable of different ECG Lead systems |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9198733B2 (en) * | 2008-04-29 | 2015-12-01 | Virginia Tech Intellectual Properties, Inc. | Treatment planning for electroporation-based therapies |
-
2018
- 2018-12-18 CN CN201811551924.5A patent/CN109480827B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN2542204Y (en) * | 2002-05-16 | 2003-04-02 | 南京市玄武区中医医院 | Vector electrocardiograph |
| JP2011143152A (en) * | 2010-01-18 | 2011-07-28 | Kyoto Nishikawa:Kk | Simple negative potential detector |
| CN201912078U (en) * | 2011-01-13 | 2011-08-03 | 深圳市科曼医疗设备有限公司 | Electrocardiogram machine for synchronously outputting electrocardio and electrocardial vector |
| CN102283639A (en) * | 2011-08-31 | 2011-12-21 | 惠州Tcl移动通信有限公司 | Heartbeat detection method, device and system, heartbeat display method and mobile terminal |
| CN104902814A (en) * | 2012-10-12 | 2015-09-09 | 加利福尼亚大学董事会 | Configuration and spatial placement of frontal electrode sensors to detect physiological signals |
| CN107530019A (en) * | 2015-03-31 | 2018-01-02 | 皇家飞利浦有限公司 | Automatic detection/the classification exchanged for the ECG cable of different ECG Lead systems |
Non-Patent Citations (5)
| Title |
|---|
| Application of an Inverse-Forward Approach to Derive the 12-lead ECG from Body Surface Potential Maps;Laura R Bear，et al;《Computing in Cardiology》;20171231;第44卷;第1-4页 * |
| Multiscale Recurrence Quantification Analysis of Spatial Cardiac Vectorcardiogram Signals;Hui Yang;《IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING》;20110228;第58卷(第2期);第339-347页 * |
| Patient-Specific Identification of Optimal Ubiquitous Electrocardiogram (U-ECG) Placement Using a Three-Dimensional Model of Cardiac Electrophysiology;Ki Moo Lim，et al;《IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING》;20130131;第60卷(第1期);第245-249页 * |
| Theoretical Evaluation of the McFee and Frank Vectorcardiographic Lead Systems Using a Numerical Inhomogeneous Torso Model;GUY E. MAILLOUX，et al;《IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING》;19820531;第29卷(第5期);第322-332页 * |
| Vickrey-Clarke-Groves for privacy-preserving collaborative classification;Anastasia Panoui,et al;《Proceedings of the 2013 Federated Conference on Computer Science and Information Systems》;20131231;第123-128页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109480827A (en) | 2019-03-19 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109480827B (en) | Vector electrocardiogram classification method and device | |
| US11344243B2 (en) | Artificial intelligence self-learning-based static electrocardiography analysis method and apparatus | |
| CN109480825B (en) | Electrocardio data processing method and device | |
| CN106214145B (en) | Electrocardiogram classification method based on deep learning algorithm | |
| Ramanathan et al. | Noninvasive electrocardiographic imaging (ECGI): application of the generalized minimal residual (GMRes) method | |
| CN106815570B (en) | Electrocardiosignal ST-T segment identification method based on dynamic pattern identification | |
| CN110769750A (en) | Electroanatomical mapping tool prompted by activation waveform | |
| DE112016005886T5 (en) | Computational localization of fibrillation sources | |
| CN106725426A (en) | A kind of method and system of electrocardiosignal classification | |
| CN114901143A (en) | System and method for reduced-lead electrocardiography diagnosis using deep neural networks and rule-based systems | |
| CN105142500A (en) | Composite singularity mapping | |
| CN108420425A (en) | Wearable dynamic maternal fetus electrocardioscanner | |
| CN106063704B (en) | A kind of QRS wave starting and terminal point localization method based on regularization least square recurrence learning | |
| JPH10211180A (en) | Method for automatically diagnosing heart function | |
| CN112971800A (en) | Deep learning-based electrocardio premature beat signal processing method, module and device | |
| Bergquist et al. | Reconstruction of cardiac position using body surface potentials | |
| Prakash et al. | An automated diagnosis system for cardiac arrhythmia classification | |
| CN111528833B (en) | Rapid identification and processing method and system for electrocardiosignals | |
| Afkhami et al. | ECG based detection of left ventricular hypertrophy using higher order statistics | |
| Bear et al. | Accuracy of lead removal vs linear interpolation in non-invasive electrocardiographic imaging (ECGI) | |
| Llamedo et al. | Analysis of 12-lead classification models for ECG classification | |
| CN112438735B (en) | Electrocardiogram P wave detection method, analysis device and storage medium | |
| Nagel et al. | Quantification of interpatient 12-lead ECG variabilities within a healthy cohort | |
| JP4402298B2 (en) | System for determining the state of the patient's heart | |
| CN112603328A (en) | Electrocardiogram wave band marking system based on deep learning |
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 |