CN109770891A - ECG signal preprocessing method and preprocessing device - Google Patents
ECG signal preprocessing method and preprocessing device Download PDFInfo
- Publication number
- CN109770891A CN109770891A CN201910095703.XA CN201910095703A CN109770891A CN 109770891 A CN109770891 A CN 109770891A CN 201910095703 A CN201910095703 A CN 201910095703A CN 109770891 A CN109770891 A CN 109770891A
- Authority
- CN
- China
- Prior art keywords
- electrocardiosignal
- period
- dimensional vector
- signal processing
- ecg signal
- 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
Landscapes
- Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)
Abstract
The present invention relates to ECG detecting Classification and Identification fields, specifically provide a kind of ECG signal processing method, comprising: S1, the electrocardiosignal that number format is obtained by the period；S2, the electrocardiosignal in each period is normalized respectively, keeps the minimum value of all electrocardiosignals identical, and maximum value is also identical；S3, the electrocardiosignal in each period is saved as into one-dimensional vector respectively, the element in the one-dimensional vector is the value after the normalization obtained by step S2；S4, the one-dimensional vector group for obtaining several by step S3 are combined into bivector.Several continuous electrocardiosignal groups are combined into one group of data by the present invention, if this group of data are inputted neural network, are conducive to computer and the related information between continuous multiple electrocardiosignals is compared and is analyzed.
Description
Technical field
The present invention relates to ECG detecting Classification and Identification field more particularly to a kind of ECG signal processing method and pretreatments
Device.
Background technique
Arrhythmia cordis, i.e. cardiac arrhythmia are a kind of main indications for reflecting heart organic disease.Currently, for analyzing
The main means of arrhythmia cordis are electrocardiograms.Since the form of expression of arrhythmia cordis is complex, if only by existing
Electrocardiogram analyzes arrhythmia cordis, it is necessary in conjunction with the rich experiences of clinician, otherwise it is easy to appear erroneous judgement or
Judge the situation of inaccuracy.
In order to simple and accurately judge the type of arrhythmia cordis, to accurately confirm illness, emerge in large numbers
It is largely based on the technical computer-aided diagnosis means of computer neural network self study out.But current area of computer aided
In diagnostic method, electrocardiogram is all regarded as a time series to consider, uses Recognition with Recurrent Neural Network popular at present
(RNN:Recurrent Neural Network) or convolutional neural networks (CNN:Convolutional Neural
Network) technology carries out the classification and Detection of the arrhythmia cordis based on electrocardiogram.
Timing dependence of the electrocardiosignal as time series is utilized in RNN technology.This method passes through some features
The technologies such as the technology of extraction such as wavelet transformation (make extract feature vector in this way can lose frequency domain information), CNN network into
Row feature extraction will extract obtained feature as the input of RNN model, the result of cardiac diagnosis exported by RNN model.But
It is that this method constructed is not the model of " end-to-end ", this " two-part " algorithm cannot be guaranteed training pattern global optimum.
CNN technology equally exists some limiting factors.Since CNN technology consolidates input data (referring mainly to data format)
Restricted, the electrocardio CNN modelling technique that is used to diagnose of all at present " end-to-end " can not be in view of the timing of electrocardiosignal
Correlative character, it is therefore desirable to ECG data be pre-processed, if dealt with improperly, largely will lead to nerve net
Network model can not carry out extensive application, practical cannot be applied in clinical diagnosis.Meanwhile often considering in preprocessing process
The problem of uncertainty to electrocardiosignal length and the electrocardiosignal containing much noise are difficult to, therefore pretreated
Increasing in journey reduces noise, extracts the step of feature vector, to remove the unrelated noise in part, still, corresponding, the meter
The performance superiority and inferiority of calculation machine aided diagnosis method also depends greatly on the performance of Denoising Algorithm and extracts feature vector quality (packet
The method for including selected characteristic vector), therefore become algorithm solution with the closely bound up data preprocessing method of the extraction of feature vector
The key for electric disease surveillance problem of being determined.
To sum up, most of data preprocessing method need to spend it is a large amount of calculate cost, and But most of algorithms is relied on
Specific hyper parameter needs are manually adjusted according to data, cause them not carry out directly, easily and quickly clinical real-time
Monitoring.
On the other hand, partial data preprocess method use be unfavorable for understand and explain mathematical measure to electrocardiosignal into
It has gone pretreatment, the physical signal that is appreciated that, can be explained originally is made to be converted into unintelligible, unaccountable data.Doctor and
Sufferer can only obtain whether suffer from the diagnostic result of certain disease, and originally equally independently data were explained and managed without the image of Buddha
Solution, for status at this stage, is unfavorable for the communication of the state of an illness made a definite diagnosis between doctors and patients.
Summary of the invention
To solve the above problems, the present invention proposes a kind of ECG signal processing method, include the following steps:
S1, the electrocardiosignal that number format is obtained by the period；
S2, the electrocardiosignal in each period is normalized respectively, makes the minimum value of all electrocardiosignals
It is identical, and maximum value is also identical；
S3, the electrocardiosignal in each period is saved as into one-dimensional vector respectively, the element in the one-dimensional vector is
Value after the normalization obtained by step S2；
S4, the one-dimensional vector group for obtaining several by step S3 are combined into bivector.
Further, in the ECG signal processing method, in step S1, a period electrocardio is obtained as follows
Signal:
S11, the R wave site for marking each cycle；
S12, a selected R wave site count the current R wave site and previous R wave site as current R wave site respectively
Between data amount check a and the current R wave site and latter R wave site between data amount check b；
S13, centered on current R wave site, take forwardA data, take backwardA data, synthesize current period
Electrocardiosignal.
Further, in the ECG signal processing method, to the electrocardiosignal in each period in step S2
Processing step include:
S21, the voltage peak-to-peak value for calculating this period electrocardiosignal；
S22, the Voltage Peak peak averaging is divided into 2nGrade, wherein n is natural number, and accordingly by minimum voltage
Value is denoted as 0, and maximum voltage value is denoted as 2n- 1, remaining sample amplitude when reproduced value is normalized into 0- (2 in proportionn- 1) numerical value between
And it is handled by rounding.
Further, in the ECG signal processing method, in step S3, for element number deficiency predetermined constant
The one-dimensional vector of m supplies element number in a manner of zero padding.
Further, in the ECG signal processing method, when zero padding before the head element of the one-dimensional vector and
Synchronous zero padding, the error for allowing 1 element occur after tail element.
Further, in the ECG signal processing method, with 5 one-dimensional vectors for one group in step S4,
A bivector is formed in such a way that sequence combines.
Further, in the ECG signal processing method, further includes: the bivector is inputted neural network
Engine is to analyze illness.
Further, in the ECG signal processing method, further includes: the bivector is shown as gray scale
Figure, in order to which human eye tentatively identifies.
Another object of the present invention is to provide a kind of ECG signal processing devices, comprising:
Signal acquisition module, for acquiring the voltage value of the electrocardiosignal；
The electrocardiosignal in each period is normalized respectively, obtains one-dimensional vector for signal normalization module；
Signal fills module, with one-dimensional vector described in null filling, keeps the element number of all one-dimensional vectors unified；
The one-dimensional vector of several signal filling module outputs is arranged successively, is combined into one by signal recombination module
Bivector.
Further, the ECG signal processing device further include: the signal is recombinated mould by gray scale display module
Each element in the bivector of block output is regarded as gray value, so that the bivector is shown as grayscale image.
Compared with prior art, the present invention proposes a kind of completely new technical solution, the skill for ECG signal processing
Art scheme has abandoned the existing thinking analyzed on the basis of voltage waveform, but will survey obtained electrocardio by the period
Signal normalization is to same numberical range, to make different cycles electrocardiosignal that can more intuitively be compared to each other.Meanwhile the present invention
Several continuous electrocardiosignal groups are combined into one group of data, if this group of data are inputted neural network, are conducive to computer pair
Related information between continuous multiple electrocardiosignals is compared and analyzes.
Also, under the premise of guarantee neural network same parameter, Method of Data Organization of the invention is more efficient, no
It needs to obtain the feature vector for characterizing ecg wave form feature by complicated calculating, also not need using being extremely difficult to instruct
Experienced recurrent neural network (RNN:Recurrent Neural Network) model comes while considering the " morphology of ecg wave form
Characteristic information " and " contextual information ".
Further, the present invention normalizes to electrocardiosignal comprising in the range of 256 values of series, so as to by normalizing
Data after change are shown as grayscale image.Similar existing electrocardiogram, from grayscale image, doctor also can recognize that a part of heart work
The information of work can solve the problems, such as data interpretation to a certain extent.
Detailed description of the invention
Fig. 1 is a kind of electrocardiosignal schematic diagram of simplification；
Fig. 2 is the original electro-cardiologic signals acquired in one embodiment of the invention；
Fig. 3 is the schematic diagram being labeled in one embodiment of the invention to original electro-cardiologic signals；
Fig. 4 is the result that the mark in one embodiment of the invention according to Fig.3, is split electrocardiosignal by the period；
Fig. 5 is the electrocardiosignal shown in the form of grayscale image in one embodiment of the invention；
Fig. 6 is the signal that several electrocardiosignals shown in fig. 5 are superposed to a width grayscale image in one embodiment of the invention.
Specific embodiment
To be clearer and more comprehensible the purpose of the present invention, feature, a specific embodiment of the invention is made with reference to the accompanying drawing
Further instruction.However, the present invention can be realized with different forms, it should not be to be confined to the embodiment described.And
In the case where not conflicting, the features in the embodiments and the embodiments of the present application allow to be combined with each other or replace.In conjunction with below
Illustrate, advantages and features of the invention will become apparent from.
It should be noted that attached drawing is all made of very simplified form and using non-accurate ratio, only to convenient, bright
The purpose of the embodiment of the present invention is aided in illustrating clearly.
Statement is also needed, the purpose of number of steps is convenient for reference in the present invention, and non-limiting sequencing.It is right
It will be illustrated in the step of need to emphasizing sequence individually, text with special text.
Computer-aided diagnosis arrhythmia cordis generally comprises four steps: ecg signal acquiring, ECG signal processing, the heart
The feature of electrocardiosignal is analyzed in the feature extraction of electric signal.Wherein, the acquisition of electrocardiosignal is the base of all diagnosis
Plinth, but include many noises in the electrocardiosignal of collected human body, so generally requiring to the electrocardiosignal comprising noise
It is pre-processed, to remove noise jamming.The feature extraction of electrocardiosignal is technological means common at present, and passes through this hair
The pretreated data of bright technical solution can directly input in neural network model, not need to extract feature.
It is the signal of the electrocardiogram of a cycle obtained by electrocardiogram acquisition instrument shown in Fig. 1.In figure, horizontal axis is
Time, the longitudinal axis are voltage value, and P, Q, R, S, T, the U marked is the general main Wave crest and wave troughs several to ecg wave form of industry
Mark point.The length and peak-to-peak value of the electrocardiosignal in each period all will not be duplicate, that is, P, Q for being marked in figure,
R, the voltage value of the characteristic points such as S, T, U and the time of appearance all change in each period.Absolute voltage value can be best table
The feature of electrocardio disease is levied, but data contain decimal, digit passes through normalized far more than normalization integer
The more convenient related algorithm processing of electrocardiosignal, so that the processing speed of electrocardio algorithm is faster.Therefore by the voltage in each period
Value, which normalizes in same numberical range, can be conducive to do correlation analysis to several continuous waveforms, effective without losing
Information.
The present invention is based on above-mentioned thinkings, propose a kind of ECG signal processing method, mainly include the following steps:
S1, the electrocardiosignal that number format is obtained by the period.Use general digital electrocardiograph.
S2, the electrocardiosignal in each period is normalized respectively, makes the minimum value of all electrocardiosignals
It is identical, and maximum value is also identical.That is, not by with clear physical significance but each periodic voltage peak-to-peak value
Unified to one of identical electrocardiosignal and physical significance can be in the identical numberical ranges of corresponding and each period peak-to-peak value.It will
Data normalization, which is conducive to electrocardio real time monitoring algorithm, can more quickly handle electrocardiogram (ECG) data, electrocardiogram (ECG) data is facilitated to handle,
Cardioelectric monitor related algorithm is disposed on the mobile apparatus.
S3, the electrocardiosignal in each period is saved as into one-dimensional vector respectively, the element in the one-dimensional vector is
Value after the normalization obtained by step S2.Since neural network often has the requirement of format to the data of input, one
In a preferred embodiment, the one-dimensional vector for element number less than 512 supplies element number in a manner of zero padding, zero padding
Position can not influence the accurate of judgement in data header in data trailer, as long as matching with neural network model yet
Property.
S4, the one-dimensional vector group for obtaining several by step S3 are combined into bivector.Combined mode is by one
One-dimensional vector is regarded as a line in bivector, several one-dimensional vectors are arranged by the sequencing of the acquisition of electrocardiosignal, from
And constitute a bivector.
Specifically, following steps can be also subdivided into step S1, it is subsequent to obtain the electrocardiosignal as unit of the period
Data processing, during Illnesses Diagnoses carried out by basic unit of the electrocardiogram (ECG) data in a period:
S11, the R wave site for marking each cycle, that is, the point of R shown in Fig. 1.
S12, a selected R wave site count the current R wave site and previous R wave site as current R wave site respectively
Between data amount check a and the current R wave site and latter R wave site between data amount check b, can referring to Fig. 3 in show
Meaning.
S13, centered on current R wave site, take forwardA data, take backwardA data, synthesize current period
Electrocardiosignal.That is, the data amount check of current period isIt is a, as shown in Figure 4.
Specifically, including: to the processing step of the electrocardiosignal in each period in step S2
S21, the voltage peak-to-peak value for calculating this period electrocardiosignal；
S22, the Voltage Peak peak averaging is divided into 2nGrade, wherein n is natural number, and accordingly by minimum voltage
Value is denoted as 0, and maximum voltage value is denoted as 2n- 1, remaining sample amplitude when reproduced value is normalized into 0- (2 in proportionn- 1) numerical value between
And it is handled by rounding.In general, maximum voltage value can be denoted as to arbitrary number, but handled for convenience of subsequent algorithm, and
Signal after guaranteeing normalization can retain the effective information of diagnosis, and ceiling voltage is usually denoted as 2n-1.Specifically, will be described
Voltage Peak peak averaging is divided into 256 grades, i.e. n=8, and minimum voltage value is denoted as 0 accordingly, and maximum voltage value is denoted as 255,
Remaining person analogizes.The reason of those skilled in the art can be understanding, is divided into 256 grades is being deposited using a byte in this present embodiment
Data are stored up, if using double byte storing data in other embodiments, the voltage peak-to-peak value can be divided into 512 completely
Grade, 1024 grades ....
Further, a kind of ECG signal processing method proposed by the present invention, can also include the following steps:
S5, the bivector obtained by step S4 is shown as grayscale image, in order to which human eye tentatively identifies.Such as back
Described in scape technology, since neural network analysis is a kind of means of mathematics, one-to-one close can not be established with physical phenomenon
System, therefore " interpretation " is lacked by the judging result that neural network obtains.However in medical practice, either doctor is gone back
It is sufferer, all necessary not only for a conclusion, but prefers to obtain an explanation, that is, conclusion is obtained based on what criterion
's.Preprocess method provided by the invention is corresponding with grayscale image by the data for inputting neural network, gives such as Fig. 4 and Fig. 5
Shown in an intuitive, visual figure, and the figure and original ecg wave form, there are corresponding relationship, doctor can basis
Existing medical image knowledge, promotes the use of in grayscale image as shown in Figure 5, can explain pass through calculating to a certain extent
It is that machine auxiliary diagnosis obtains the result is that how to obtain.
Correspondingly, the invention also provides a kind of ECG signal processing devices, comprising:
Signal acquisition module, for acquiring the voltage value of the electrocardiosignal；
The electrocardiosignal in each period is normalized respectively, obtains one-dimensional vector for signal normalization module；
Signal fills module, with one-dimensional vector described in null filling, keeps the element number of all one-dimensional vectors unified；
The one-dimensional vector of several signal filling module outputs is arranged successively, is combined into one by signal recombination module
Bivector.
Further, the ECG signal processing device further includes a gray scale display module, the gray scale display module
Each element in the bivector of signal recombination module output is regarded as gray value, so that the bivector be shown
For grayscale image.
Below in conjunction with Fig. 2~Fig. 6, the process of ECG signal processing is elaborated.
As shown in Fig. 2, signal acquisition module collects multiple continuous electrocardiosignals, wherein 5 are intercepted in the present embodiment
The waveform of continuous cycles is as example.
As shown in figure 3, the segmentation in order to realize signal period, the waveforms of the present embodiment 5 continuous cycles shown in Fig. 2 into
Rower note, using the R point in ecg wave form as the central point of segmentation.Since the gap periods of ecg wave form are not completely the same
, that is to say, that the sampled data number between two R points is different, therefore, in order to reasonably divide signal period, this reality
Apply example count respectively sampled data number a between the R point of current period and previous R point and the R in preceding period point and previous R point it
Between sampled data number b.
As shown in figure 4, the present embodiment is taken forward centered on current R pointA sampled data, takes backwardA hits
According to electrocardiogram (ECG) data as current period.Also, the electrocardiosignal amplitude in multiple periods is normalized, then can be obtained
To an one-dimensional vector.In the present embodiment, the one-dimensional vector specifically: [..., 255 ..., 0 ...].Those skilled in the art can
Know, the least member in one-dimensional vector at this time is 0, greatest member 255, and the element number in each one-dimensional vector is not to the utmost
It is identical.Neural network generally requires uniform length to input data, that is to say, that if the element number of the one-dimensional vector is insufficient
M (predetermined constant) is a, then needs to be extended to m element by treaty rule.In general, can be by the member of the one-dimensional vector
Plain number is unified to arbitrary number, but subsequent algorithm is handled for convenience, and considers the sample frequency of electrocardiosignal, and
Signal after guaranteeing normalization can retain diagnosis effective information, and the present embodiment sets m=512.The present embodiment is above-mentioned one
" 0 " (zero padding) is filled in dimensional vector, so that the element number of all one-dimensional vectors is unified for 512.The method of zero padding can have
A variety of, the present embodiment is using by the way of the head of original one-dimensional vector, the synchronous filling " 0 " of tail.Specifically, it is above-mentioned it is one-dimensional to
Amount is corrected for: [0,0,0 ..., 255 ..., 0 ..., 0,0,0].Since the element number of original one-dimensional vector is indefinite, if solidification
In the head of original one-dimensional vector, the synchronous filling " 0 " of tail, then be likely to occur the case where being unable to reach 512 elements.It at this time can be with
Selection fills one zero in head or tail portion less, the accuracy of the deviation not impact analysis of a data.
It is the one-dimensional vector shown in the form of grayscale image shown in Fig. 5.The data in totally 5 periods in figure, respectively and in Fig. 4
5 periods data it is corresponding.
The grayscale image in 5 periods in Fig. 5 is arranged successively and (is equivalent to the one-dimensional vector group being combined into bivector)
Then obtain grayscale image shown in fig. 6.According to the grey scale change situation of the grayscale image, doctor can preliminary analysis electrocardio situation, to disease
Suffer from and explains possible illness.Bivector corresponding with the grayscale image then can be used for training neural network, alternatively, can input
The mature neural network of training, realizes computer-aided diagnosis.
Under conditions of guarantee neural network same parameter, above-mentioned ECG signal processing method and pretreatment unit
So that the data of input neural network are more efficient.Data prediction is carried out using the present invention, is not needed through complicated calculating
The feature vector for characterizing ecg wave form feature is obtained, does not need to come while examining using being extremely difficult to trained RNN model
Consider ecg wave form " morphological feature information " and " contextual information ".The real-time monitoring that thus can be realized electrocardiosignal is examined
It is disconnected, and can more be quickly obtained corresponding diagnostic result.
Further, the present invention is more easily applied to clinical practice, compared to the correlation not considered between ecg wave form
Method, the present invention can obtain the relevant information of higher level diagnosis by way of this special data organization.
Further, Method of Data Organization of the invention is not only restricted to the leading number limitation of ECG detecting, for common
All lead numbers data, can efficiently support.The data of multiple leads are constituted side by side by being once segmented
The mode of bivector can simultaneously take into account the information of all leads, the organizational form of this data, in depth network
Under the ability of the powerful pattern-recognition of model, it is able to ascend the upper limit of the algorithm of cardiac diagnosis.
Obviously, those skilled in the art can carry out various modification and variations without departing from spirit of the invention to invention
And range.If in this way, these modifications and changes of the present invention belong to the claims in the present invention and its equivalent technologies range it
Interior, then the present invention is also intended to include these modifications and variations.
Claims (10)
1. a kind of ECG signal processing method, which comprises the steps of:
S1, the electrocardiosignal that number format is obtained by the period；
S2, the electrocardiosignal in each period is normalized respectively, keeps the minimum value of all electrocardiosignals identical,
And maximum value is also identical；
S3, the electrocardiosignal in each period is saved as into one-dimensional vector respectively, the element in the one-dimensional vector is to pass through
Value after the normalization that step S2 is obtained；
S4, the one-dimensional vector group for obtaining several by step S3 are combined into bivector.
2. ECG signal processing method as described in claim 1, which is characterized in that in step S1, obtain as follows
One period electrocardiosignal:
S11, the R wave site for marking each cycle；
S12, a selected R wave site count between the current R wave site and previous R wave site respectively as current R wave site
Data amount check a and the current R wave site and latter R wave site between data amount check b；
S13, centered on current R wave site, take forwardA data, take backwardA data synthesize the heart of current period
Electric signal.
3. ECG signal processing method as claimed in claim 1 or 2, which is characterized in that each period in step S2
The processing step of the electrocardiosignal includes:
S21, the voltage peak-to-peak value for calculating this period electrocardiosignal；
S22, the Voltage Peak peak averaging is divided into 2nGrade, wherein n is natural number, and is accordingly denoted as minimum voltage value
0, maximum voltage value is denoted as 2n- 1, remaining sample amplitude when reproduced value is normalized into 0- (2 in proportionn- 1) numerical value and process between
Rounding processing.
4. ECG signal processing method as described in claim 1, which is characterized in that in step S3, not for element number
The one-dimensional vector of sufficient predetermined constant m supplies element number in a manner of zero padding.
5. ECG signal processing method as claimed in claim 4, which is characterized in that in the head of the one-dimensional vector when zero padding
With synchronous zero padding after tail element before element, the error for allowing 1 element occur.
6. the ECG signal processing method as described in claim 1 or 4 or 5, which is characterized in that in step S4 described in 5
One-dimensional vector is one group, forms a bivector in such a way that sequence combines.
7. ECG signal processing method as described in claim 1, which is characterized in that further include: the bivector is defeated
Enter neural network model to analyze illness.
8. ECG signal processing method as described in claim 1, which is characterized in that further include: the bivector is shown
It is shown as grayscale image, in order to which human eye tentatively identifies.
9. a kind of ECG signal processing device characterized by comprising
Signal acquisition module, for acquiring the voltage value of the electrocardiosignal；
The electrocardiosignal in each period is normalized respectively, obtains one-dimensional vector for signal normalization module；
Signal fills module, with one-dimensional vector described in null filling, keeps the element number of all one-dimensional vectors unified；
The one-dimensional vector of several signal filling module outputs is arranged successively, is combined into a two dimension by signal recombination module
Vector.
10. ECG signal processing device as described in claim 1, which is characterized in that further include:
Each element in the bivector of signal recombination module output is regarded as gray value by gray scale display module, thus
The bivector is shown as grayscale image.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910095703.XA CN109770891B (en) | 2019-01-31 | 2019-01-31 | Electrocardiosignal preprocessing method and preprocessing device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910095703.XA CN109770891B (en) | 2019-01-31 | 2019-01-31 | Electrocardiosignal preprocessing method and preprocessing device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109770891A true CN109770891A (en) | 2019-05-21 |
| CN109770891B CN109770891B (en) | 2022-04-29 |
Family
ID=66503066
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910095703.XA Active CN109770891B (en) | 2019-01-31 | 2019-01-31 | Electrocardiosignal preprocessing method and preprocessing device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109770891B (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111091092A (en) * | 2019-12-16 | 2020-05-01 | 桂林电子科技大学 | A Data Augmentation Method Based on Optimizing Convolutional Neural Network Classification |
| CN113349790A (en) * | 2020-10-16 | 2021-09-07 | 山东大学 | Heart beat classification method and system of 2D convolutional neural network based on S-shaped reconstruction |
| CN113609336A (en) * | 2021-08-16 | 2021-11-05 | 山东大学 | XML document standardization method and system for multiple electrocardiosignal storage |
| CN117556337A (en) * | 2023-03-07 | 2024-02-13 | 上海理工大学 | A damage identification method for underground pipelines based on distributed optical fiber sensing system and residual network |
Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1725978A (en) * | 2002-12-13 | 2006-01-25 | 皇家飞利浦电子股份有限公司 | System and method for processing a series of image frames representing a cardiac cycle |
| CN103126669A (en) * | 2012-12-21 | 2013-06-05 | 深圳市理邦精密仪器股份有限公司 | A method for improving the accuracy of waveform drawing and a drawing method and device using the same |
| CN103610457A (en) * | 2013-11-19 | 2014-03-05 | 深圳先进技术研究院 | Electrocardiosignal processing method and electrocardiosignal processing system |
| CN104367317A (en) * | 2014-10-15 | 2015-02-25 | 北京理工大学 | Electrocardiogram electrocardiosignal classification method with multi-scale characteristics combined |
| KR101524226B1 (en) * | 2013-08-19 | 2015-05-29 | 광운대학교 산학협력단 | Method of classifing heart condition using neural network, and apparatus thereof |
| CN105105743A (en) * | 2015-08-21 | 2015-12-02 | 山东省计算中心（国家超级计算济南中心） | Intelligent electrocardiogram diagnosis method based on deep neural network |
| US20160135704A1 (en) * | 2014-11-18 | 2016-05-19 | Siemens Medical Solutions Usa, Inc. | Matrix-Based Patient Signal Analysis |
| US20170112401A1 (en) * | 2015-10-27 | 2017-04-27 | CardioLogs Technologies | Automatic method to delineate or categorize an electrocardiogram |
| CN107837082A (en) * | 2017-11-27 | 2018-03-27 | 乐普（北京）医疗器械股份有限公司 | Electrocardiogram automatic analysis method and device based on artificial intelligence self study |
| CN107981858A (en) * | 2017-11-27 | 2018-05-04 | 乐普（北京）医疗器械股份有限公司 | Electrocardiogram heartbeat automatic recognition classification method based on artificial intelligence |
| CN108962393A (en) * | 2018-05-12 | 2018-12-07 | 鲁东大学 | Automatic arrhythmia analysis method based on compression figure neural network |
| CN108968951A (en) * | 2018-08-15 | 2018-12-11 | 武汉中旗生物医疗电子有限公司 | Electrocardiogram detecting method, apparatus and system |
| CN109171712A (en) * | 2018-09-28 | 2019-01-11 | 东软集团股份有限公司 | Auricular fibrillation recognition methods, device, equipment and computer readable storage medium |
-
2019
- 2019-01-31 CN CN201910095703.XA patent/CN109770891B/en active Active
Patent Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1725978A (en) * | 2002-12-13 | 2006-01-25 | 皇家飞利浦电子股份有限公司 | System and method for processing a series of image frames representing a cardiac cycle |
| CN103126669A (en) * | 2012-12-21 | 2013-06-05 | 深圳市理邦精密仪器股份有限公司 | A method for improving the accuracy of waveform drawing and a drawing method and device using the same |
| KR101524226B1 (en) * | 2013-08-19 | 2015-05-29 | 광운대학교 산학협력단 | Method of classifing heart condition using neural network, and apparatus thereof |
| CN103610457A (en) * | 2013-11-19 | 2014-03-05 | 深圳先进技术研究院 | Electrocardiosignal processing method and electrocardiosignal processing system |
| CN104367317A (en) * | 2014-10-15 | 2015-02-25 | 北京理工大学 | Electrocardiogram electrocardiosignal classification method with multi-scale characteristics combined |
| US20160135704A1 (en) * | 2014-11-18 | 2016-05-19 | Siemens Medical Solutions Usa, Inc. | Matrix-Based Patient Signal Analysis |
| CN105105743A (en) * | 2015-08-21 | 2015-12-02 | 山东省计算中心（国家超级计算济南中心） | Intelligent electrocardiogram diagnosis method based on deep neural network |
| US20170112401A1 (en) * | 2015-10-27 | 2017-04-27 | CardioLogs Technologies | Automatic method to delineate or categorize an electrocardiogram |
| CN107837082A (en) * | 2017-11-27 | 2018-03-27 | 乐普（北京）医疗器械股份有限公司 | Electrocardiogram automatic analysis method and device based on artificial intelligence self study |
| CN107981858A (en) * | 2017-11-27 | 2018-05-04 | 乐普（北京）医疗器械股份有限公司 | Electrocardiogram heartbeat automatic recognition classification method based on artificial intelligence |
| CN108962393A (en) * | 2018-05-12 | 2018-12-07 | 鲁东大学 | Automatic arrhythmia analysis method based on compression figure neural network |
| CN108968951A (en) * | 2018-08-15 | 2018-12-11 | 武汉中旗生物医疗电子有限公司 | Electrocardiogram detecting method, apparatus and system |
| CN109171712A (en) * | 2018-09-28 | 2019-01-11 | 东软集团股份有限公司 | Auricular fibrillation recognition methods, device, equipment and computer readable storage medium |
Non-Patent Citations (3)
| Title |
|---|
| 夏灵: "《基于心电仿真模型参数解的心电逆问题研究》", 30 June 2002 * |
| 查雪帆: "结合迁移学习与深度卷积网络的心电分类研究", 《中国医学物理学杂志》 * |
| 马玉润: "ECG预处理与QRS波群检测技术研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111091092A (en) * | 2019-12-16 | 2020-05-01 | 桂林电子科技大学 | A Data Augmentation Method Based on Optimizing Convolutional Neural Network Classification |
| CN113349790A (en) * | 2020-10-16 | 2021-09-07 | 山东大学 | Heart beat classification method and system of 2D convolutional neural network based on S-shaped reconstruction |
| CN113609336A (en) * | 2021-08-16 | 2021-11-05 | 山东大学 | XML document standardization method and system for multiple electrocardiosignal storage |
| CN113609336B (en) * | 2021-08-16 | 2024-03-26 | 山东大学 | XML document standardization method and system for storing multiple electrocardiosignals |
| CN117556337A (en) * | 2023-03-07 | 2024-02-13 | 上海理工大学 | A damage identification method for underground pipelines based on distributed optical fiber sensing system and residual network |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109770891B (en) | 2022-04-29 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11564612B2 (en) | Automatic recognition and classification method for electrocardiogram heartbeat based on artificial intelligence | |
| CN109770891A (en) | ECG signal preprocessing method and preprocessing device | |
| CN101152081B (en) | Method and apparatus with reduced electrode system specific ecg interpretation | |
| US9161705B2 (en) | Method and device for early detection of heart attack | |
| WO2021071688A1 (en) | Systems and methods for reduced lead electrocardiogram diagnosis using deep neural networks and rule-based systems | |
| CN107361760B (en) | Magnetocardiogram instrument diagnosis system and magnetocardiogram instrument using the same | |
| EP2224346B1 (en) | Ecg signal analysis tool | |
| WO2021071646A1 (en) | Systems and methods for electrocardiogram diagnosis using deep neural networks and rule-based systems | |
| CN108573227A (en) | ECG data quality evaluating method and device | |
| CN109288508A (en) | A kind of pressure value intelligent measurement method based on CRNN-BP | |
| CN109846474A (en) | The processing method and processing device of electrocardiogram, the long-range processing method of electrocardiogram and system | |
| CN113080996A (en) | Electrocardiogram analysis method and device based on target detection | |
| CN114494798A (en) | A kind of electrocardiogram artifact confirmation method, terminal device and storage medium | |
| CN118021316B (en) | Dynamic electrocardiographic data processing method and device, electronic equipment and storage medium | |
| JP6715397B1 (en) | Data processing device, data processing method and program | |
| CN118285790A (en) | Exercise function rehabilitation prediction method based on multi-mode asynchronous data and related equipment | |
| CN113171102A (en) | ECG data classification method based on continuous deep learning | |
| CN101596125A (en) | Health information display system with visual display, method and interface thereof | |
| CN117100276B (en) | Cardiac function detection system, computer storage medium and terminal | |
| CN115969388A (en) | Doppler radar heartbeat detection method based on multi-scale residual shrinkage network | |
| US20040234508A1 (en) | Arrangement with artificial neurons for describing the transmission behavior of a nerve cell to be excited and method for determining the transmission behavior of a nerve cell to be excited using artificial neurons | |
| CN119523499B (en) | A method and device for evaluating electroencephalogram data quality | |
| CN118021315B (en) | Dynamic electrocardiographic data processing method and device, electronic equipment and storage medium | |
| US20050261874A1 (en) | Method and arrangement and computer programme with programme code means and computer programme products for the analysis of neuronal activities in neuronal areas | |
| BAMMARA et al. | Classification of ECG signals using 1D-2D transformation and convolutional neural networks (CNN) |
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 |