CN108968951A - Electrocardiogram detecting method, apparatus and system - Google Patents
Electrocardiogram detecting method, apparatus and system Download PDFInfo
- Publication number
- CN108968951A CN108968951A CN201810928372.9A CN201810928372A CN108968951A CN 108968951 A CN108968951 A CN 108968951A CN 201810928372 A CN201810928372 A CN 201810928372A CN 108968951 A CN108968951 A CN 108968951A
- Authority
- CN
- China
- Prior art keywords
- electrocardiogram
- subclass
- model
- detection model
- detected
- 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.)
- Granted
Links
Classifications
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
- A61B5/316—Modalities, i.e. specific diagnostic methods
- A61B5/318—Heart-related electrical modalities, e.g. electrocardiography [ECG]
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
- A61B5/316—Modalities, i.e. specific diagnostic methods
-
- 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
-
- 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)
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Pathology (AREA)
- Biomedical Technology (AREA)
- Heart & Thoracic Surgery (AREA)
- Medical Informatics (AREA)
- Molecular Biology (AREA)
- Surgery (AREA)
- Animal Behavior & Ethology (AREA)
- Biophysics (AREA)
- Public Health (AREA)
- Veterinary Medicine (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physiology (AREA)
- Psychiatry (AREA)
- Signal Processing (AREA)
- Evolutionary Computation (AREA)
- Fuzzy Systems (AREA)
- Mathematical Physics (AREA)
- Cardiology (AREA)
- Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)
Abstract
The application provides a kind of electrocardiogram detecting method, apparatus and system, and the electrocardiogram to be detected received is inputted SVM model, obtains output signal distribution and the signal label of electrocardiogram to be detected；When signal label is that abnormal signal marks, it is distributed to obtain target image according to abnormal signal label and output signal, and target image is inputted big classification detection model, obtains the target symptom classification of electrocardiogram to be detected.Detect the symptom subclass of electrocardiogram to be detected respectively by the first subclass detection model and the second subclass detection model, comprehensive detection is carried out to electrocardiogram to be detected further according to obtained two testing results and above-mentioned target symptom classification, finally to determine the symptom subclass of electrocardiogram to be detected.In this way, on the one hand may not necessarily artificial adjusting parameter, on the other hand improve the accuracy rate of detection.
Description
Technical field
This application involves medical image process fields, in particular to a kind of electrocardiogram detecting method, device and are
System.
Background technique
Electrocardiogram is mainly used for reacting the electric ignition process of heart, is that doctor carries out cardiac work up and one of diagnosis important
Clinical means.Electrocardiogram complexity itself is strong, the otherness of different race, sexes, the people at age under various pathologic conditions
It is very big.In actual clinical diagnosis, it is usually auxiliary with machine testing result, combines itself clinical experience to the heart by doctor
Electrograph carries out judgement identification.It in this case, usually can be excessively since the knowledge of doctor is professional and the deficiency of experience accumulation
The testing result that machine provides is relied on, accuracy rate is had certain limitations, the erroneous judgement to abnormal electrocardiographic pattern is easy to cause.
In the prior art, neural network is generallyd use to classify to electrocardiosignal, with find electrocardiosignal time domain,
The different characteristic information for being included in frequency domain, the type of feature extraction is more, and the accuracy rate of classification is higher.Common learning method
It is based on the heart marked in the arrhythmia cordis database of MIT and claps type identification, the identification of normal abnormal signal or for a certain special
Different pathology electrocardiosignal carries out identification classification, however, be limited to the sample size of electrocardiosignal, the universality of this method compared with
Difference.
In the prior art usually by manually extracting characteristic information, and the characteristic information type manually extracted is limited, can shadow
Ring the accuracy classified to electrocardiosignal.
Summary of the invention
In view of this, the application's is designed to provide a kind of electrocardiogram detecting method, apparatus and system, at least partly
Ground improves the above problem.
In order to achieve the above object, the embodiment of the present application adopts the following technical scheme that
In a first aspect, the embodiment of the present application provides a kind of electrocardiogram detecting method, which comprises
The electrocardiogram to be detected received is inputted into SVM model, obtains the output signal distribution of the electrocardiogram to be detected
And signal label；
When the signal label is that abnormal signal marks, it is distributed according to abnormal signal label and the output signal
Target image is obtained, and the target image is inputted into big classification detection model, to identify corresponding to the electrocardiogram to be detected
Target symptom classification, wherein the big classification detection model be CNN model；
Input the mesh of the electrocardiogram to be detected respectively to the first subclass detection model and the second subclass detection model
Lead signals are marked, respectively obtain two probability groups, each probability group includes that the electrocardiogram to be detected belongs to each symptom subclass
Probability, wherein the first subclass detection model be CNN model, the second subclass detection model be LSTM model；
Based on comprehensive detection model, judged jointly according to described two probability groups and the target symptom classification described to be checked
Symptom subclass belonging to thought-read electrograph, wherein the comprehensive descision model is Attention model.
Second aspect, the embodiment of the present application also provide a kind of ECG tester, and described device includes:
Abnormal judgment module obtains the electrocardio to be detected for the electrocardiogram to be detected received to be inputted SVM model
The output signal of figure is distributed and signal label；
Big classification detection module is used for when the signal label is that abnormal signal marks, according to the abnormal signal mark
Note, the output signal are distributed to obtain target image, and the target image is inputted big classification detection model, described in identification
Target symptom classification corresponding to electrocardiogram to be detected, wherein the big classification detection model is CNN model；
Subclass detection module, for inputting institute respectively to the first subclass detection model and the second subclass detection model
The target lead signals for stating electrocardiogram to be detected, respectively obtain two probability groups, and each probability group includes the electrocardio to be detected
Figure belongs to the probability of each symptom subclass, wherein the first subclass detection model is CNN model, second subclass
Detection model is LSTM model；
Comprehensive detection module, for being based on comprehensive detection model, according to described two probability groups and the target symptom class
Symptom subclass belonging to the electrocardiogram to be detected is not judged jointly, wherein the comprehensive descision model is Attention mould
Type.
The third aspect, the embodiment of the present application also provide a kind of electrocardiogram detection system, which includes being in communication with each other connection
Data server, deep learning server and application server；
The application server is by network and user terminal communication, the electrocardio to be detected that the user terminal is sent
Figure is transmitted to the deep learning server；
The deep learning server includes processor and machine readable storage medium, is deposited on the machine readable storage medium
Machine readable instructions are contained, which, which is performed, promotes the processor to realize the heart provided by the embodiments of the present application
Electrograph detection method is detected with the electrocardiogram to be detected sent to the application server；
The data server is used to provide training sample to the deep learning server, for deep learning clothes
Business device carries out deep learning.
In terms of existing technologies, the embodiment of the present application has the advantages that
A kind of electrocardiogram detecting method provided by the embodiments of the present application, apparatus and system, the electrocardio to be detected that will be received
Figure input SVM model obtains output signal distribution and the signal label of electrocardiogram to be detected；When signal label is abnormal signal mark
It clocks, is distributed to obtain target image according to abnormal signal label and output signal, and target image is inputted big classification and detects mould
Type obtains the target symptom classification of electrocardiogram to be detected.Pass through the first subclass detection model and the second subclass detection model
The symptom subclass for detecting electrocardiogram to be detected respectively, further according to obtained two testing results and above-mentioned target symptom classification pair
Electrocardiogram to be detected carries out comprehensive detection, finally to determine the symptom subclass of electrocardiogram to be detected.In this way, on the one hand can not
Must artificial adjusting parameter, on the other hand improve the accuracy rate of detection.
Detailed description of the invention
Technical solution in ord to more clearly illustrate embodiments of the present application, below will be to needed in the embodiment attached
Figure is briefly described, it should be understood that the following drawings illustrates only some embodiments of the application, therefore is not construed as pair
The restriction of range for those of ordinary skill in the art without creative efforts, can also be according to this
A little attached drawings obtain other relevant attached drawings.
Fig. 1 is a kind of connection schematic diagram of electrocardiogram detection system provided by the embodiments of the present application；
Fig. 2 is a kind of block diagram of deep learning server provided by the embodiments of the present application；
Fig. 3 is a kind of flow diagram of electrocardiogram detecting method provided by the embodiments of the present application；
Fig. 4 is a kind of the functional block diagram of ECG tester provided by the embodiments of the present application.
Icon: 10- electrocardiogram detection system；11- deep learning server；111- memory；112- processor；113- is logical
Believe unit；12- data server；13- application server；20- user terminal；30- ECG tester；31- judges extremely
Module；The big classification detection module of 32-；33- subclass detection module；34- comprehensive detection module.
Specific embodiment
To keep the purposes, technical schemes and advantages of the embodiment of the present application clearer, below in conjunction with the embodiment of the present application
In attached drawing, the technical scheme in the embodiment of the application is clearly and completely described, it is clear that described embodiment is
Some embodiments of the present application, instead of all the embodiments.The application being usually described and illustrated herein in the accompanying drawings is implemented
The component of example can be arranged and be designed with a variety of different configurations.
Therefore, the detailed description of the embodiments herein provided in the accompanying drawings is not intended to limit below claimed
Scope of the present application, but be merely representative of the selected embodiment of the application.Based on the embodiment in the application, this field is common
Technical staff's every other embodiment obtained without creative efforts belongs to the model of the application protection
It encloses.
It should also be noted that similar label and letter indicate similar terms in following attached drawing, therefore, once a certain Xiang Yi
It is defined in a attached drawing, does not then need that it is further defined and explained in subsequent attached drawing.
Fig. 1 is please referred to, is a kind of connection schematic diagram of electrocardiogram detection system 10 provided by the embodiments of the present application.The heart
Electrograph detection system 10 includes the deep learning server 11, data server 12 and application server 13 for being in communication with each other connection,
Wherein, deep learning server 11, data server 12 and application server 13 can be located in same local area network.
Application server 13 can be web application server, logical as the cardiac electricity detecting system 10 and external equipment
The interface of letter can be communicated to connect by Ethernet and user terminal 20.In the embodiment of the present application, user terminal 20 can be
Personal computer (PersonalComputer, PC), smart phone, tablet computer etc. arbitrarily have data processing function and communication
The equipment of function.
When implementing, user can will need the electrocardiogram to be detected detected to be sent to the application by user terminal 20
The electrocardiogram to be detected is sent to the deep learning server 11, the depth by server 13, the application server 13
Learning server 11 is spent to be used to carry out the electrocardiogram to be detected by electrocardiogram detecting method provided by the embodiments of the present application
Detection.The data server 12 is for providing training sample to the deep learning server 11, to carry out deep learning.
Optionally, in the embodiment of the present application, the data server 12 can be an individual server, can also be with
It is the data server cluster of more data servers composition.Accordingly, the deep learning server 12 can be individually
One server is also possible to the deep learning server cluster of more deep learning servers composition.Wherein, the data clothes
Business device cluster and the deep learning server cluster can be realized by the way of distributed type assemblies, to improve data access
Speed, and carry out redundancy backup.
Incorporated by reference to the box signal for referring to Fig. 2, Fig. 2 being a kind of deep learning server 11 provided by the embodiments of the present application
Figure.The deep learning server 11 includes ECG tester 30, memory 111, processor 112 and communication unit 113.
The memory 111, processor 112 and each element of communication unit 113 directly or indirectly electrically connect between each other
It connects, to realize the transmission or interaction of data.For example, these elements can pass through one or more communication bus or signal between each other
Line, which is realized, to be electrically connected.The ECG tester 30 includes at least one can be in the form of software or firmware (firmware)
Be stored in the memory 111 or be solidificated in the deep learning server 11 operating system (operating system,
OS the software function module in).The processor 112 is used to execute the executable module stored in the memory 111, such as
Software function module and computer program etc. included by the ECG tester 30.
Wherein, the memory 111 may be, but not limited to, random access memory (Random Access
Memory, RAM), read-only memory (Read Only Memory, ROM), programmable read only memory (Programmable
Read-Only Memory, PROM), erasable read-only memory (Erasable Programmable Read-Only
Memory, EPROM), electricallyerasable ROM (EEROM) (Electric Erasable Programmable Read-Only
Memory, EEPROM) etc..
The processor 112 can be a kind of IC chip, the processing capacity with signal.Above-mentioned processor can
To be general processor, including central processing unit (Central Processing Unit, CPU), network processing unit (Network
Processor, NP) etc.；It can also be digital signal processor (DSP), specific integrated circuit (ASIC), field-programmable gate array
Arrange (FPGA) either other programmable logic device, discrete gate or transistor logic, discrete hardware components.It may be implemented
Or disclosed each method, step and logic diagram in execution the embodiment of the present application.General processor can be microprocessor
Or the processor is also possible to any conventional processor etc..
The communication unit 113 is used for the communication connection established between the deep learning server 11 and other equipment,
Such as it can be by local area network foundation and the communication connection between the data server 12 and the application server 13, again
Such as, the communication connection between the user terminal 20 can be established by Ethernet.
It should be appreciated that structure shown in Fig. 2 is only to illustrate, the deep learning server 11 can also include than Fig. 2 institute
Show more or fewer components, or with configuration entirely different shown in Fig. 2.In addition, each component shown in Fig. 2 can be with
Software, hardware or combinations thereof are realized.
It is worth noting that, the data server 12, application server 13 and user in the embodiment of the present application are whole herein
Connection relationship between the included component and component such as end 20, can be similar with the deep learning server 11, herein
It repeats no more.
Referring to figure 3., Fig. 3 is a kind of flow diagram of electrocardiogram detecting method provided by the embodiments of the present application, described
Method can be applied to deep learning server 11 shown in Fig. 2, below will be to the method includes each steps to carry out in detail
It illustrates.
The electrocardiogram to be detected received is inputted SVM model, obtains the output of the electrocardiogram to be detected by step S31
Signal distributions and signal label.
Wherein, the electrocardiogram to be detected inputs the SVM after being converted into digitized electrocardiosignal to be detected
(SupportVectorMachine, support vector machines) model, the SVM model be used for judge input electrocardiosignal whether be
Abnormal signal.
Wherein, the electrocardiosignal to be detected include 12 lead signals, 12 lead signals be respectively as follows: I, II,
III, aVR, aVL, aVF, V1, V2, V3, V4, V5 and V6.It, can be defeated as a width picture using 12 lead signals when implementing
Enter, 12 lead signals can also be handled respectively is 12 width pictures, retells the 12 width picture together as input letter
Number；Or wherein 6 lead signals of 12 lead signals are saved as into a width picture, in addition 6 lead signals save as one
Width picture retells two width pictures and is used as input signal together；Or using 3 × 4 mode, i.e., every 3 lead signals are saved
For a width picture, 4 width pictures are obtained, then regard the 4 width picture as input signal together.The present embodiment is without limitation.
In practical applications, different sexes, the people at age, the waveform morphology of normal ECG is different, such as: 0-
3 years old babies, 3-6 years old and 6-12 years old children, since its heart is won anatomical features, electrocardiogram and adult have significantly
It is different.In traditional normal ECG criterion of identification, according to heart rate, P-R interphase, the QRS time limit, QTc interphase, QRS wave electric axis,
The different shape and variation tendency of P wave, QRS wave and T wave in 12 lead signals carry out comprehensive descision, to determine electrocardiogram
It is whether abnormal.Accordingly, in the embodiment of the present application, first judge whether electrocardiosignal to be detected is abnormal signal.
When implementing, a series of calculating operation is carried out to the electrocardiosignal to be detected, it is reflected from the input space
It is mapped to output space, and by a series of processing such as projection, transformation, mappings, the calculated result and output label in space will be exported
Continuous iteration, backpropagation calculating, amendment are carried out, a series of data distribution groups of the different output labels of characterizations are finally obtained
It closes.By using the SVM model, the minimum of risk and fiducial range may be implemented, thus reach statistical sample amount compared with
In the case where few, the purpose of good statistical law can be also obtained.
In the present embodiment, the SVM model is mapped the data of the electrocardiosignal to be detected by a kernel function K
It is linear inseparable in luv space to solve the problems, such as into higher dimensional space.The objective function of the SVM model can be with are as follows:Wherein, φ indicates the feature space mapping of input signal x to f, i.e. dual form converts.?
In the present embodiment, kernel function needs to meet: K (x, z)=<φ (x) φ (z)>, and inner product calculating is carried out in feature space, can be with
It is specific as follows using gaussian kernel function:
In the present embodiment, the value of the signal label of the SVM model output can be 1 or -1, wherein 1 is normal signal
Label, -1 marks for abnormal signal.In other words, when the signal label of output is 1, indicate that the electrocardiosignal to be detected is positive
Regular signal label；When the signal label of output is -1, indicate that the electrocardiosignal to be detected is abnormal signal.
Step S32, when the signal label is that abnormal signal marks, according to abnormal signal label and the output
Signal distributions obtain target image, and the target image is inputted big classification detection model, to identify the electrocardio to be detected
The corresponding target symptom classification of figure, wherein the big classification detection model is CNN model.
Wherein, the target image can be obtained by following process: be believed abnormal signal label and the output
Number distribution carry out pixel value normalized, can specifically normalize to 0-255, then obtained normalization data is spliced to
The top of one standard cardioelectric figure obtains the target image that size is 200 × 1000.
In practical applications, usually exist normal electrocardio, arrhythmia cordis, intraventricular block, atrioventricular hypertrophy, ST sections it is different
Often, six symptom classifications of myocardial infarction, wherein each symptom classification includes multiple symptom subclass again.In the present embodiment, needle
To each symptom classification, the highest at least ten symptom subclass of the frequency of occurrences in the symptom classification is determined, for identified
Each symptom subclass, obtains at least 10000 parts of adult's electrocardiograms of the symptom subclass, and by the wave in every part of electrocardiogram
Shape signal is converted to corresponding digital signal, obtains corresponding electrocardiosignal sample.In this way, above-mentioned each electrocardio reports the corresponding heart
Electric signal sample forms a training sample database.In this way, in the training sample database include 51 symptom subclass, each 10000 parts
Data.
In the present embodiment, can by the training sample database to the SVM model, the big classification detection model and
Other subsequent models are trained.
During digitized processing, it is assumed that each symptom classification in the training sample database includes 10 symptom subclasses
Not, then processing can be numbered to above-mentioned classification, wherein corresponding number of normal electrocardio is 1000, remaining five symptom class
Symptom subclass in not is respectively 2001-2010,3001-3010,4001-4010,5001-5010,6001-6010.It is corresponding
Gender can be indicated that, for example, male is 1, women 0 by ground with 0 and 1.Age can directly be inputted with numerical value.
Optionally, in the present embodiment, the big classification detection model is trained according to the training sample database
Step may include following sub-step:
First, for each training sample, mark when the abnormal signal for obtaining the training sample by the SVM model and
When output signal is distributed, the normalization that is normalized, and will obtain is distributed to abnormal signal label and output signal
Data are spliced to above a standard cardioelectric figure and obtain the first image；And
Second, obtain the patient information of the training sample, and by the electrocardiosignal sample obtained according to the training sample and
The patient information joint generates the second image, wherein the electrocardiosignal sample includes 12 leads letter of the training sample
Number.
Wherein, the size of second image can be 1000 × 1000 pixels.
Third is trained the big classification detection model by the first image and second image, and
The parameter that the big classification detection model is adjusted in training process makes the big classification detection model that each trained sample be recognized accurately
This target symptom classification.
In the present embodiment, the big classification detection model may include multiple groups be made of convolutional layer and down-sampled layer
It closes, the multiple combination is sequentially connected, and the output of the down-sampled layer in previous combination is the input of the convolutional layer in next combination.
Wherein, the multiple combination is respectively the first combination, the second combination, third combination, the 4th combination and the 5th combination.
In view of the parameter sharing of CNN (Convolutional Neural Network, convolutional neural networks) model is special
Property, and observing this higher specificity of the abundanter accuracy of characteristic value, the convolutional layer C1 in first combination can wrap
64 convolution kernels are included, which can use Sobel operator, and the size of each convolution kernel is 9*9.Wherein, using sobel
The edge feature of image can be enhanced in operator.By convolutional layer C1,200 different characteristics of available input picture, totally 64
A characteristic pattern, the size of each characteristic pattern are 1192*992.
Optionally, in convolutional layer C1, each neuron can have 100 connection weight parameters.
Optionally, the convolutional layer C1 can be using linear amending unit PRuLU function as activation primitive, the function
Are as follows:
In the case, due to negative semiaxis slopeOn-fixed is constant, therefore PReLU convergence rate is faster, calculates
Journey are as follows:
Down-sampled layer S1 in first combination can be by the way of 4 × 4 average pond to the convolutional layer C1
The characteristic pattern of output carries out down-sampling, and by down-sampled layer S1, the size of characteristic pattern be can reduce as 298*248.
Optionally, in the present embodiment, the convolutional layer C2 in the second combination may include 128 using sobel operator
Convolution kernel, the size of each convolution kernel are 5*5, and the convolutional layer C2 can be using tanh function as activation primitive.Accordingly,
Down-sampled layer S2 in second combination can be used for carrying out 2 × 2 down-sampling, available by the down-sampled layer S2
Size is the characteristic pattern of 147*122.
Optionally, in the present embodiment, third combination in convolutional layer C3 can be used for the output to down-sampled layer S2 into
Row process of convolution, convolutional layer C3 include the convolution kernel that 256 sizes are 6*5, the size of the characteristic pattern of the convolutional layer C3 output
For 142*118.Down-sampled layer S3 in third combination can be used for carrying out 2 × 2 average pond, the characteristic pattern size of output
For 71*59.
Optionally, in the present embodiment, the convolutional layer C4 in the 4th combination is used to carry out convolution, convolution to down-sampled layer S3
Layer C4 includes the convolution kernel that 512 sizes are 6*6, and the size of the characteristic pattern of convolutional layer C4 output is 66*54.In 4th combination
Down-sampled layer S4 is used to carry out 2 × 2 average pond, is the characteristic pattern of 33*27 with output size.
Optionally, in the present embodiment, the convolutional layer C5 in the 5th combination includes the convolution kernel that 512 sizes are 4*4, with
Output size is the characteristic pattern of 30*24；Down-sampled layer S5 in 5th combination is used to carry out 2 × 2 average pond, big to export
The small characteristic pattern for 15*12.Wherein, the five down-sampled layers S5 down-sampled layer of most end (that is, in the multiple combination)
It is connected with full articulamentum and output layer in turn, wherein the full articulamentum can be activated using sigmoid function, the output
Layer can be softmax layers.
In the present embodiment, the output layer may include multiple output nodes, and each output node corresponds to a kind of illness
Classification.In a specific embodiment, the output layer may include 7 output nodes, lose respectively with normal electrocardio, the rhythm of the heart
Often, intraventricular block, atrioventricular hypertrophy, ST sections of abnormal, myocardial infarctions and 7 kinds of situations of other classifications are corresponding.
In the present embodiment, the output result of the big classification detection module can be inputted into CNN and LSTM respectively
(LongShort-TermMemory, shot and long term memory network) two kinds of deep learning network, to carry out symptom subclass
Detection.
Step S33 inputs the thought-read to be checked to the first subclass detection model and the second subclass detection model respectively
The target lead signals of electrograph, respectively obtain two probability groups, and each probability group includes that the electrocardiogram to be detected belongs to each disease
Levy the probability of subclass, wherein the first subclass detection model is CNN model, and the second subclass detection model is
LSTM model.
In a specific embodiment, the target lead signals can be 12 lead signals.It is specific at another
In embodiment, since in 12 lead signals, III, aVL and aVF lead are the linear combination of I, II lead, therefore can be only
Using I, II, V1, V2, V3, V4, V5 and V6, totally 8 lead signals as target lead signals are input to the inspection of above-mentioned two subclass
It surveys in model.
Optionally, step S33 may include following sub-step:
Using the target lead signals as input signal, preset duration t is carried out to the input signal for sample rate with 500
Sampling；
Obtained sampled data and the patient information are combined into the target data of t × 5000, by the target data point
The first subclass detection model and the second subclass detection model are not inputted.
Wherein, the preset duration t can be 10 seconds (s), and the unit of the sample rate is Hz.
In addition, in the present embodiment, due to the signal of each lead reflect be Heart tissue electro-physiological signals,
Therefore, the conduction current potential of heartbeat belongs to highly relevant signal on same time point, therefore, it is possible to using 10 × 10 sizes
Convolution kernel carries out feature extraction to electrocardiosignal.
Specifically, the first subclass detection model can be using 3 convolutional layers, 3 pond layers, 1 output layer and 1 packet
The full articulamentum output of 1024 neurons is included, finally the probability of i.e. exportable each subclass (symptom subclass).Wherein, described
3 convolutional layers and 3 pond layer specific structures are as shown in the table:
| Layer name | Convolution kernel size | Convolution kernel number | Convolution step-length |
| Convolutional layer 1 | 10×10 | 64 | 1×2 |
| Pond layer 1 | 1×1 | - | - |
| Convolutional layer 2 | 1×10 | 128 | 1×5 |
| Pond layer 2 | 1×4 | - | - |
| Convolutional layer 3 | 10×10 | 512 | 1×10 |
| Pond layer 3 | 1×2 | - | - |
About the second subclass detection model, it is contemplated that electrocardiosignal in a lead signals, each heart beat signal it
Between point there is time continuity, reflect heart during primary beating, current potential is transmitted to atrium, the heart from sinoatrial node output
The synthesis current potential at other positions of room exports, thus, electrocardiosignal has serial correlation.Based on this, LSTM model can be used
The second subclass detection model is realized, because LSTM model is a kind of special RNN
(RecurrentNeuralNetwork) network, Dependence Problem when can well solve long.
In the present embodiment, LSTM model may include 4 layers of structure, totally 10 steps, and sequence length is in individual data
5000 points, hidden layer includes 200 neurons, need to carry out 100000 training altogether.
In the present embodiment, LSTM model mainly includes three doors and a memory unit (cell), wherein described three
Door is respectively forgetgate, inputgate and outputgate.The forgetgate can be used for through door control unit pair
Cell addition and deletion information.By forgetgate, it can choose and determine whether information passes through comprising sigmoid nerve
Network layer and a pairs of multiplication operation, output are the numerical value between 0 to 1, and the numerical value allows for indicating
By information number, wherein 0 indicate do not allow to pass through completely, 1 indicate allow to pass through completely.Inputgate passes through
Sigmoid come determine which value for updating, tanh layers for generating new candidate value Ct, its candidate as current layer generation
Value may be added in cellstate, and the value that this part generates can be combined for updating.Outputgate is for certainly
Determine the output of LSTM model, obtains an initial output by sigmoid layers first, reuse tanh function and scale Ct value
To between -1 and 1, then the output that obtains with sigmoid is by being multiplied, to obtain the output of model.
Step S34 is based on comprehensive detection model, is judged jointly according to described two probability groups and the target symptom classification
Symptom subclass belonging to the electrocardiogram to be detected, wherein the comprehensive descision model is Attention model.
Disease is being carried out by CNN model (the first subclass detection model) and LSTM model (the second subclass detection model)
After levying subclass detection, the projective transformation of multilayer is carried out to electrocardiosignal to be detected, has been calculated, and has obtained each disease
Levy the corresponding output probability of subclass.However, it has been investigated that, the quantity of symptom subclass is more, final calculated
The possible very little of most probable value is not enough to prove be some subclass, thus, in the embodiment of the present application, pass through a comprehensive inspection
The probability of the first subclass detection model described in models coupling and the second subclass detection model output is surveyed, and with described big
Classification detection model output target symptom classification as priori knowledge and the target lead signals detect jointly described in
Detect symptom subclass belonging to electrocardiosignal.
In the case where the output for stating multiple models before addition, then combine the electrocardiosignal to be detected, can mention significantly
The accuracy of high final classification result, the weight between the neuron of comprehensive detection model can also solve well.
As shown in figure 4, being a kind of functional block diagram of ECG tester 30 provided by the embodiments of the present application.It is described
ECG tester 30 includes abnormal judgment module 31, big classification detection module 32, subclass detection module 33 and comprehensive inspection
Survey module 34.
Wherein, the abnormal judgment module 31 is used to the electrocardiogram to be detected received inputting SVM model, obtains described
The output signal of electrocardiogram to be detected is distributed and signal label.
In the present embodiment, the description as described in the abnormal judgment module 31 is specifically referred to step S31 shown in Fig. 3
Detailed description, i.e. step S31 can be executed by the abnormal judgment module 31.
The big classification detection module 32 is used for when the signal label is that abnormal signal marks, according to the abnormal letter
Labelled notation, the output signal are distributed to obtain target image, and the target image is inputted big classification detection model, with identification
Target symptom classification corresponding to the electrocardiogram to be detected, wherein the big classification detection model is CNN model.
In the present embodiment, the description as described in the big classification detection module 32 is specifically referred to step S32 shown in Fig. 3
Detailed description, i.e. step S32 can execute by the big classification detection module 32.
The subclass detection module 33 is used to distinguish to the first subclass detection model and the second subclass detection model
The target lead signals for inputting the electrocardiogram to be detected, respectively obtain two probability groups, and each probability group includes described to be checked
Thought-read electrograph belongs to the probability of each symptom subclass, wherein the first subclass detection model be CNN model, described second
Subclass detection model is LSTM model.
In the present embodiment, the description as described in the subclass detection module 33 is specifically referred to step S33 shown in Fig. 3
Detailed description, i.e. step S33 can execute by the subclass detection module 33.
The comprehensive detection module 34 is used to be based on comprehensive detection model, according to described two probability groups and target disease
Sign classification judges symptom subclass belonging to the electrocardiogram to be detected jointly, wherein the comprehensive descision model is
Attention model.
In the present embodiment, description specifically can be with reference pair step shown in Fig. 3 as described in the comprehensive detection module 34
The detailed description namely step S34 of S34 can be executed by the comprehensive detection module 34.
In conclusion a kind of electrocardiogram detecting method provided by the embodiments of the present application, apparatus and system, will receive to
It detects electrocardiogram and inputs SVM model, obtain output signal distribution and the signal label of electrocardiogram to be detected；When signal label is different
When regular signal marks, it is distributed to obtain target image according to abnormal signal label and output signal, and target image is inputted major class
Other detection model obtains the target symptom classification of electrocardiogram to be detected.Pass through the first subclass detection model and the second subclass
Detection model detects the symptom subclass of electrocardiogram to be detected respectively, further according to two obtained testing results and above-mentioned target disease
It levies classification and comprehensive detection is carried out to electrocardiogram to be detected, finally to determine the symptom subclass of electrocardiogram to be detected.In this way, a side
Face may not necessarily artificial adjusting parameter, on the other hand improve the accuracy rate of detection.
In embodiment provided herein, it should be understood that disclosed device and method, it can also be by other
Mode realize.The apparatus embodiments described above are merely exemplary, for example, the flow chart and block diagram in attached drawing are shown
According to device, the architectural framework in the cards of method and computer program product, function of multiple embodiments of the application
And operation.In this regard, each box in flowchart or block diagram can represent one of a module, section or code
Point, a part of the module, section or code includes one or more for implementing the specified logical function executable
Instruction.It should also be noted that function marked in the box can also be attached to be different from some implementations as replacement
The sequence marked in figure occurs.For example, two continuous boxes can actually be basically executed in parallel, they sometimes may be used
To execute in the opposite order, this depends on the function involved.It is also noted that each of block diagram and or flow chart
The combination of box in box and block diagram and or flow chart can be based on the defined function of execution or the dedicated of movement
The system of hardware is realized, or can be realized using a combination of dedicated hardware and computer instructions.
In addition, each functional module in each embodiment of the application can integrate one independent portion of formation together
Point, it is also possible to modules individualism, an independent part can also be integrated to form with two or more modules.
It, can be with if the function is realized and when sold or used as an independent product in the form of software function module
It is stored in a computer readable storage medium.Based on this understanding, the technical solution of the application is substantially in other words
The part of the part that contributes to existing technology or the technical solution can be embodied in the form of software products, the meter
Calculation machine software product is stored in a storage medium, including some instructions are used so that a computer equipment (can be a
People's computer, server or network equipment etc.) execute each embodiment the method for the application all or part of the steps.
And storage medium above-mentioned includes: that USB flash disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), arbitrary access are deposited
The various media that can store program code such as reservoir (RAM, Random Access Memory), magnetic or disk.
It should be noted that, in this document, relational terms such as first and second and the like are used merely to a reality
Body or operation are distinguished with another entity or operation, are deposited without necessarily requiring or implying between these entities or operation
In any actual relationship or order or sequence.Moreover, the terms "include", "comprise" or its any other variant are intended to
Non-exclusive inclusion, so that the process, method, article or equipment including a series of elements is not only wanted including those
Element, but also including other elements that are not explicitly listed, or further include for this process, method, article or equipment
Intrinsic element.In the absence of more restrictions, the element limited by sentence "including a ...", it is not excluded that
There is also other identical elements in process, method, article or equipment including the element.
The above, the only specific embodiment of the application, but the protection scope of the application is not limited thereto, it is any
Those familiar with the art within the technical scope of the present application, can easily think of the change or the replacement, and should all contain
Lid is within the scope of protection of this application.Therefore, the protection scope of the application should be subject to the protection scope in claims.
Claims (10)
1. a kind of electrocardiogram detecting method, which is characterized in that the described method includes:
The electrocardiogram to be detected received is inputted into SVM model, obtains output signal distribution and the letter of the electrocardiogram to be detected
Number label；
When the signal label is that abnormal signal marks, it is distributed to obtain according to abnormal signal label and the output signal
Target image, and the target image is inputted into big classification detection model, to identify mesh corresponding to the electrocardiogram to be detected
Mark symptom classification, wherein the big classification detection model is CNN model；
The target for inputting the electrocardiogram to be detected respectively to the first subclass detection model and the second subclass detection model is led
Join signal, respectively obtains two probability groups, each probability group includes that the electrocardiogram to be detected belongs to the general of each symptom subclass
Rate, wherein the first subclass detection model is CNN model, and the second subclass detection model is LSTM model；
Based on comprehensive detection model, the thought-read to be checked is judged jointly according to described two probability groups and the target symptom classification
Symptom subclass belonging to electrograph, wherein the comprehensive descision model is Attention model.
2. electrocardiogram detecting method according to claim 1, which is characterized in that marked according to the abnormal signal, is described
Output signal is distributed to obtain target image, comprising:
The normalized of pixel value, the normalization that will be obtained are carried out to abnormal signal label and output signal distribution
Data are spliced to above a standard cardioelectric figure, obtain the target image.
3. electrocardiogram detecting method according to claim 1 or 2, which is characterized in that the method also includes:
Training sample database is established, includes normal electrocardio, arrhythmia cordis, intraventricular block, chamber fertilizer in the training sample database
Greatly, ST sections of abnormal, six symptom classifications of myocardial infarction samples, each symptom classification includes the highest at least ten of the frequency of occurrences
The sample of symptom subclass, wherein each symptom subclass has at least 10000 parts of adult's electrocardiograms；
Mould is detected to the SVM model, the big classification detection model, first subclass according to the training sample database
Type, the second subclass detection model and the comprehensive detection model are trained.
4. electrocardiogram detecting method according to claim 3, which is characterized in that according to the training sample database to described big
Classification detection model is trained, comprising:
For each training sample, when the abnormal signal label and output signal for obtaining the training sample by the SVM model
When distribution, abnormal signal label and output signal distribution are normalized, and obtained normalization data is spliced
The first image is obtained above to a standard cardioelectric figure；And
The patient information of the training sample is obtained, and the electrocardiosignal sample obtained according to the training sample and the patient are believed
Breath joint generates the second image, wherein the electrocardiosignal sample includes 12 lead signals of the training sample；
The big classification detection model is trained by the first image and second image, and in the training process
The parameter for adjusting the big classification detection model makes the big classification detection model that the target disease of each training sample be recognized accurately
Levy classification.
5. electrocardiogram detecting method according to claim 1 or 2, which is characterized in that
The objective function of the SVM model are as follows:
The kernel function of the SVM model is gaussian kernel function:
6. electrocardiogram detecting method according to claim 1 or 2, which is characterized in that the big classification detection model includes
The multiple combinations being made of convolutional layer and down-sampled layer, the multiple combination are sequentially connected, the down-sampled layer in previous combination
Output is the input of the convolutional layer in next combination, and the down-sampled layer of most end is connected with full articulamentum and output layer, institute in turn
Stating output layer includes multiple output nodes, and each output node corresponds to a kind of symptom classification；The target symptom classification is by such as
Lower step identification:
Obtain multiple probability values that the multiple output node exports respectively；
Using symptom classification corresponding to the maximum value in the multiple probability value as the target symptom classification.
7. electrocardiogram detecting method according to claim 1 or 2, which is characterized in that the target lead signals include institute
State I, II, V1, V2, V3, V4, V5 and V6 in electrocardiogram to be detected totally 8 lead signals.
8. electrocardiogram detecting method according to claim 7, which is characterized in that the first subclass detection model and second
Subclass detection model inputs the target lead signals of the electrocardiogram to be detected respectively, comprising:
Using the target lead signals as input signal, preset duration t is carried out to the input signal for sample rate with 500 and is adopted
Sample；
Obtained sampled data and the patient information are combined into the target data of t × 5000, which distinguished defeated
Enter the first subclass detection model and the second subclass detection model.
9. a kind of ECG tester, which is characterized in that described device includes:
Abnormal judgment module obtains the electrocardiogram to be detected for the electrocardiogram to be detected received to be inputted SVM model
Output signal distribution and signal label；
Big classification detection module, for being marked according to the abnormal signal, institute when the signal label is that abnormal signal marks
It states output signal to be distributed to obtain target image, and the target image is inputted into big classification detection model, it is described to be checked to identify
Target symptom classification corresponding to thought-read electrograph, wherein the big classification detection model is CNN model；
Subclass detection module, for the first subclass detection model and the second subclass detection model input respectively described in
The target lead signals for detecting electrocardiogram, respectively obtain two probability groups, each probability group includes the electrocardiogram category to be detected
In the probability of each symptom subclass, wherein the first subclass detection model is CNN model, the second subclass detection
Model is LSTM model；
Comprehensive detection module, it is total according to described two probability groups and the target symptom classification for being based on comprehensive detection model
It is same to judge symptom subclass belonging to the electrocardiogram to be detected, wherein the comprehensive descision model is Attention model.
10. a kind of electrocardiogram detection system, which is characterized in that the system includes the data server for being in communication with each other connection, depth
Learning server and application server；
By network and user terminal communication, the electrocardiogram to be detected that the user terminal is sent turns the application server
Issue the deep learning server；
The deep learning server includes processor and machine readable storage medium, is stored on the machine readable storage medium
Machine readable instructions, which, which is performed, promotes the processor to realize described in any one of claim 1-8
Electrocardiogram detecting method, detected with the electrocardiogram to be detected sent to the application server；
The data server is used to provide training sample to the deep learning server, for the deep learning server
Carry out deep learning.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810928372.9A CN108968951B (en) | 2018-08-15 | 2018-08-15 | Electrocardiogram detection method, device and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810928372.9A CN108968951B (en) | 2018-08-15 | 2018-08-15 | Electrocardiogram detection method, device and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108968951A true CN108968951A (en) | 2018-12-11 |
| CN108968951B CN108968951B (en) | 2021-06-22 |
Family
ID=64553247
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810928372.9A Active CN108968951B (en) | 2018-08-15 | 2018-08-15 | Electrocardiogram detection method, device and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108968951B (en) |
Cited By (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109480825A (en) * | 2018-12-13 | 2019-03-19 | 武汉中旗生物医疗电子有限公司 | The processing method and processing device of electrocardiogram (ECG) data |
| CN109700434A (en) * | 2018-12-28 | 2019-05-03 | 武汉中旗生物医疗电子有限公司 | Meta-learning model training method, system and equipment based on electrocardiographic diagnosis |
| CN109770891A (en) * | 2019-01-31 | 2019-05-21 | 上海交通大学 | ECG signal processing method and pretreatment unit |
| CN110507311A (en) * | 2019-08-27 | 2019-11-29 | 中科麦迪人工智能研究院(苏州)有限公司 | A kind of ecg analysis method, apparatus, equipment and medium based on image information |
| CN110507299A (en) * | 2019-04-11 | 2019-11-29 | 研和智能科技(杭州)有限公司 | Heart rate signal detection device and method |
| CN110717416A (en) * | 2019-09-24 | 2020-01-21 | 上海数创医疗科技有限公司 | Neural network training method for ST segment classification recognition based on feature selection |
| CN110991561A (en) * | 2019-12-20 | 2020-04-10 | 山东大学齐鲁医院 | Method and system for identifying images of endoscope in lower digestive tract |
| CN111436926A (en) * | 2020-04-03 | 2020-07-24 | 山东省人工智能研究院 | Atrial fibrillation signal detection method based on statistical characteristics and convolution cyclic neural network |
| CN111657921A (en) * | 2020-07-01 | 2020-09-15 | 杭州脉流科技有限公司 | Real-time electrocardio abnormality monitoring method and device, computer equipment and storage medium |
| CN111882559A (en) * | 2020-01-20 | 2020-11-03 | 深圳数字生命研究院 | ECG signal acquisition method and device, storage medium and electronic device |
| CN112022141A (en) * | 2020-08-07 | 2020-12-04 | 上海联影智能医疗科技有限公司 | Electrocardiosignal category detection method and device and storage medium |
| CN112562846A (en) * | 2019-09-25 | 2021-03-26 | 中国联合网络通信集团有限公司 | Animal disease diagnosis device |
| CN112971796A (en) * | 2021-02-09 | 2021-06-18 | 中山大学附属第三医院（中山大学肝脏病医院） | Myocardial infarction prediction device, system and storage medium |
| CN114224355A (en) * | 2020-09-07 | 2022-03-25 | 中国移动通信有限公司研究院 | Electrocardiosignal classification training method, classification method, device and storage medium |
| CN115115038A (en) * | 2022-08-30 | 2022-09-27 | 合肥心之声健康科技有限公司 | Model construction method based on single lead electrocardiosignal and gender identification method |
| TWI808915B (en) * | 2022-11-22 | 2023-07-11 | 國立勤益科技大學 | Automatic localization method of st change events by using dynamic electrocardiography and deep learning model |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO1997017893A1 (en) * | 1995-11-13 | 1997-05-22 | Heart Rhythm Technologies, Inc. | System and method for analyzing electrogram waveforms |
| US6731973B2 (en) * | 2001-06-12 | 2004-05-04 | Ge Medical Systems Information Technologies, Inc. | Method and apparatus for processing physiological data |
| CN1539372A (en) * | 2003-10-24 | 2004-10-27 | �Ϻ���ͨ��ѧ | Method and device for early diagnosis of heart disease basaed on high-frequency waveform of cardiograph |
| CN1997313A (en) * | 2004-04-05 | 2007-07-11 | 惠普开发有限公司 | Cardiac diagnostic system and method |
| CN101129261A (en) * | 2007-02-09 | 2008-02-27 | 北京中医药大学 | Device and method for acquiring and recognizing pulsation information and tongue inspection information |
| CN106845098A (en) * | 2017-01-17 | 2017-06-13 | 天津艾登科技有限公司 | A kind of implementation method of the medical diagnosis on disease packet based on decision Tree algorithms |
| CN107408144A (en) * | 2014-11-14 | 2017-11-28 | Zoll医疗公司 | Medical precursor event estimation |
| CN107463772A (en) * | 2017-07-20 | 2017-12-12 | 广州慧扬健康科技有限公司 | The constructing system of multi-C vector spectrum of disease |
| CN108090509A (en) * | 2017-12-13 | 2018-05-29 | 四川大学 | A kind of adaptive electrocardiogram sorting technique of data length |
-
2018
- 2018-08-15 CN CN201810928372.9A patent/CN108968951B/en active Active
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO1997017893A1 (en) * | 1995-11-13 | 1997-05-22 | Heart Rhythm Technologies, Inc. | System and method for analyzing electrogram waveforms |
| US6731973B2 (en) * | 2001-06-12 | 2004-05-04 | Ge Medical Systems Information Technologies, Inc. | Method and apparatus for processing physiological data |
| CN1539372A (en) * | 2003-10-24 | 2004-10-27 | �Ϻ���ͨ��ѧ | Method and device for early diagnosis of heart disease basaed on high-frequency waveform of cardiograph |
| CN1997313A (en) * | 2004-04-05 | 2007-07-11 | 惠普开发有限公司 | Cardiac diagnostic system and method |
| CN101129261A (en) * | 2007-02-09 | 2008-02-27 | 北京中医药大学 | Device and method for acquiring and recognizing pulsation information and tongue inspection information |
| CN107408144A (en) * | 2014-11-14 | 2017-11-28 | Zoll医疗公司 | Medical precursor event estimation |
| CN106845098A (en) * | 2017-01-17 | 2017-06-13 | 天津艾登科技有限公司 | A kind of implementation method of the medical diagnosis on disease packet based on decision Tree algorithms |
| CN107463772A (en) * | 2017-07-20 | 2017-12-12 | 广州慧扬健康科技有限公司 | The constructing system of multi-C vector spectrum of disease |
| CN108090509A (en) * | 2017-12-13 | 2018-05-29 | 四川大学 | A kind of adaptive electrocardiogram sorting technique of data length |
Cited By (21)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109480825A (en) * | 2018-12-13 | 2019-03-19 | 武汉中旗生物医疗电子有限公司 | The processing method and processing device of electrocardiogram (ECG) data |
| CN109700434A (en) * | 2018-12-28 | 2019-05-03 | 武汉中旗生物医疗电子有限公司 | Meta-learning model training method, system and equipment based on electrocardiographic diagnosis |
| CN109770891A (en) * | 2019-01-31 | 2019-05-21 | 上海交通大学 | ECG signal processing method and pretreatment unit |
| CN110507299A (en) * | 2019-04-11 | 2019-11-29 | 研和智能科技(杭州)有限公司 | Heart rate signal detection device and method |
| CN110507299B (en) * | 2019-04-11 | 2022-05-13 | 研和智能科技(杭州)有限公司 | Heart rate signal detection device and method |
| CN110507311A (en) * | 2019-08-27 | 2019-11-29 | 中科麦迪人工智能研究院(苏州)有限公司 | A kind of ecg analysis method, apparatus, equipment and medium based on image information |
| CN110717416A (en) * | 2019-09-24 | 2020-01-21 | 上海数创医疗科技有限公司 | Neural network training method for ST segment classification recognition based on feature selection |
| CN110717416B (en) * | 2019-09-24 | 2021-07-09 | 上海数创医疗科技有限公司 | Neural network training method for ST segment classification recognition based on feature selection |
| CN112562846A (en) * | 2019-09-25 | 2021-03-26 | 中国联合网络通信集团有限公司 | Animal disease diagnosis device |
| CN110991561A (en) * | 2019-12-20 | 2020-04-10 | 山东大学齐鲁医院 | Method and system for identifying images of endoscope in lower digestive tract |
| CN111882559A (en) * | 2020-01-20 | 2020-11-03 | 深圳数字生命研究院 | ECG signal acquisition method and device, storage medium and electronic device |
| CN111882559B (en) * | 2020-01-20 | 2023-10-17 | 深圳数字生命研究院 | ECG signal acquisition method and device, storage medium and electronic device |
| CN111436926B (en) * | 2020-04-03 | 2021-04-20 | 山东省人工智能研究院 | Atrial fibrillation signal detection method based on statistical characteristics and convolution cyclic neural network |
| CN111436926A (en) * | 2020-04-03 | 2020-07-24 | 山东省人工智能研究院 | Atrial fibrillation signal detection method based on statistical characteristics and convolution cyclic neural network |
| CN111657921A (en) * | 2020-07-01 | 2020-09-15 | 杭州脉流科技有限公司 | Real-time electrocardio abnormality monitoring method and device, computer equipment and storage medium |
| CN112022141A (en) * | 2020-08-07 | 2020-12-04 | 上海联影智能医疗科技有限公司 | Electrocardiosignal category detection method and device and storage medium |
| CN114224355A (en) * | 2020-09-07 | 2022-03-25 | 中国移动通信有限公司研究院 | Electrocardiosignal classification training method, classification method, device and storage medium |
| CN112971796A (en) * | 2021-02-09 | 2021-06-18 | 中山大学附属第三医院（中山大学肝脏病医院） | Myocardial infarction prediction device, system and storage medium |
| CN115115038A (en) * | 2022-08-30 | 2022-09-27 | 合肥心之声健康科技有限公司 | Model construction method based on single lead electrocardiosignal and gender identification method |
| CN115115038B (en) * | 2022-08-30 | 2022-11-08 | 合肥心之声健康科技有限公司 | Model construction method based on single lead electrocardiosignal and gender identification method |
| TWI808915B (en) * | 2022-11-22 | 2023-07-11 | 國立勤益科技大學 | Automatic localization method of st change events by using dynamic electrocardiography and deep learning model |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108968951B (en) | 2021-06-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108968951A (en) | Electrocardiogram detecting method, apparatus and system | |
| Wang et al. | A high-precision arrhythmia classification method based on dual fully connected neural network | |
| Nagavelli et al. | Machine Learning Technology‐Based Heart Disease Detection Models | |
| Jahmunah et al. | Automated detection of coronary artery disease, myocardial infarction and congestive heart failure using GaborCNN model with ECG signals | |
| US20230143594A1 (en) | Systems and methods for reduced lead electrocardiogram diagnosis using deep neural networks and rule-based systems | |
| Luz et al. | ECG-based heartbeat classification for arrhythmia detection: A survey | |
| Park et al. | Hierarchical support vector machine based heartbeat classification using higher order statistics and hermite basis function | |
| Bashar et al. | Novel density poincaré plot based machine learning method to detect atrial fibrillation from premature atrial/ventricular contractions | |
| Redmond et al. | Electrocardiogram signal quality measures for unsupervised telehealth environments | |
| CN110090012A (en) | A kind of human body diseases detection method and testing product based on machine learning | |
| CN106725426A (en) | A kind of method and system of electrocardiosignal classification | |
| Yang et al. | Identification of myocardial infarction (MI) using spatio-temporal heart dynamics | |
| CN110464334A (en) | A kind of electrocardiographic abnormality detection method based on composite depth learning network | |
| JP2021186686A (en) | Local noise identification using coherent algorithm | |
| Zhang et al. | [Retracted] An ECG Heartbeat Classification Method Based on Deep Convolutional Neural Network | |
| EP4041073A1 (en) | Systems and methods for electrocardiogram diagnosis using deep neural networks and rule-based systems | |
| Wei et al. | Atrial fibrillation detection by the combination of recurrence complex network and convolution neural network | |
| Zhuang et al. | Arrhythmia diagnosis of young martial arts athletes based on deep learning for smart medical care | |
| Huang et al. | A multiview feature fusion model for heartbeat classification | |
| Wang et al. | Multiscale residual network based on channel spatial attention mechanism for multilabel ECG classification | |
| Metan et al. | Cardiovascular MRI image analysis by using the bio inspired (sand piper optimized) fully deep convolutional network (Bio-FDCN) architecture for an automated detection of cardiac disorders | |
| TWI688371B (en) | Intelligent device for atrial fibrillation signal pattern acquisition and auxiliary diagnosis | |
| CN111329467A (en) | Heart disease auxiliary detection method based on artificial intelligence | |
| Bhave et al. | Deep learning to detect left ventricular structural abnormalities in chest X-rays | |
| CN109044347A (en) | Identify electrocardio wave image junctional escape beat method, apparatus, system and electronic 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 |