CN108309284A - ECG T wave end-point detection method and device - Google Patents
ECG T wave end-point detection method and device Download PDFInfo
- Publication number
- CN108309284A CN108309284A CN201810214428.4A CN201810214428A CN108309284A CN 108309284 A CN108309284 A CN 108309284A CN 201810214428 A CN201810214428 A CN 201810214428A CN 108309284 A CN108309284 A CN 108309284A
- Authority
- CN
- China
- Prior art keywords
- wave
- electrocardiosignal
- waves
- history
- ecg
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
- A61B5/316—Modalities, i.e. specific diagnostic methods
- A61B5/318—Heart-related electrical modalities, e.g. electrocardiography [ECG]
- A61B5/346—Analysis of electrocardiograms
- A61B5/349—Detecting specific parameters of the electrocardiograph cycle
-
- 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
- A61B5/7235—Details of waveform analysis
- A61B5/725—Details of waveform analysis using specific filters therefor, e.g. Kalman or adaptive filters
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7235—Details of waveform analysis
- A61B5/7264—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems
- A61B5/7267—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems involving training the classification device
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/14—Fourier, Walsh or analogous domain transformations, e.g. Laplace, Hilbert, Karhunen-Loeve, transforms
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Health & Medical Sciences (AREA)
- Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Physics & Mathematics (AREA)
- Molecular Biology (AREA)
- Artificial Intelligence (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- General Health & Medical Sciences (AREA)
- Theoretical Computer Science (AREA)
- Heart & Thoracic Surgery (AREA)
- Animal Behavior & Ethology (AREA)
- Medical Informatics (AREA)
- Public Health (AREA)
- Veterinary Medicine (AREA)
- Pathology (AREA)
- General Physics & Mathematics (AREA)
- Surgery (AREA)
- Data Mining & Analysis (AREA)
- Cardiology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Software Systems (AREA)
- Signal Processing (AREA)
- Psychiatry (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Computational Mathematics (AREA)
- Physiology (AREA)
- Mathematical Analysis (AREA)
- Mathematical Optimization (AREA)
- Pure & Applied Mathematics (AREA)
- Algebra (AREA)
- Computing Systems (AREA)
- Databases & Information Systems (AREA)
- Computational Linguistics (AREA)
- Fuzzy Systems (AREA)
- Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)
Abstract
The present invention relates to a kind of ECG T wave end-point detection method and devices, include the following steps:Obtain T wave end point determination models;T wave end point determination models are to be carried out obtained from learning training to history T wave discrete datas based on MLP neural network models;Wherein, history T waves discrete data is to history electrocardiosignal extracts from QT databases T waves, carries out dimension-reduction treatment and obtain;Electrocardiosignal to be detected is inputted into T wave end point determination models, obtains the T wave terminal point coordinates of electrocardiosignal to be detected.In above-mentioned ECG T wave end-point detection method when obtaining T wave end point determination patterns, effectively image dimension reduction method is combined with MPL neural network models, so that obtained T wave detection model calculation amounts are smaller, when identifying T wave terminals using T wave detection models, calculation amount is small, substantially increases the efficiency of calculating.
Description
Technical field
The present invention relates to medical signals processing technology field, more particularly to a kind of ECG T wave end-point detection method and
Device.
Background technology
Electrocardiogram (electrocardiogram, abbreviation ECG or EKG) is every from body surface record heart using electrocardiograph
Electrical activity caused by one cardiac cycle changes figure, reflects the relationship between myocardial contraction and time.In electrocardiogram
The phase reflects the overall process that ventricle removes multipole between QT, and the extension of phase is that the important mark of ventricular arrhythmia is assessed in clinic between QT
Will.Therefore, the extraction of QT waves is just particularly important.
Clinically, it is very crucial step in the extraction of QT waves to the positioning of T wave terminals;However due to T wave terminals near
Signal low-frequency component is abundant, is easily mixed into noise, and T wave morphologies are easy to happen variation, therefore just very tired to the positioning of T wave terminals
It is difficult.Such as studies have shown that can generally cause the variation of T wave morphologies when ischemic class cardiomyopathy occurs, such as T waves are inverted, are two-way,
To be impacted to the detection of T wave terminals.
Currently, the detection algorithm of T wave terminals includes mainly by area, template matches, wavelet transformation, curvature, statistics mould
The methods of formula identification and neural network, these methods do not rely on merely threshold value and are handled.Wherein, by template matches,
The calculation amount of the methods of statistical-simulation spectrometry and neural network is larger, inefficient.
Invention content
Based on this, it is necessary to for current T wave terminal location algorithms when to ECG T wave end point determination calculation amount compared with
Greatly, inefficient problem provides a kind of ECG T wave end-point detection method and device.
A kind of ECG T wave end-point detection method, includes the following steps:
Obtain T wave end point determination models;The T waves end point determination model is based on MLP neural network models to history T waves
Discrete data carries out obtained from learning training;Wherein, the history T wave discrete datas are to the history electrocardio from QT databases
T waves that signal extraction goes out carry out what dimension-reduction treatment obtained;
Electrocardiosignal to be detected is inputted into the T waves end point determination model, obtains the T of the electrocardiosignal to be detected
Wave terminal point coordinate.
A kind of ECG T wave end point determination device, including:
Model acquisition module, for obtaining T wave end point determination models;The T waves end point determination model is based on MLP nerves
Network model carries out history T wave discrete datas obtained from learning training;Wherein, the history T wave discrete datas be to from
History electrocardiosignal extracts in QT databases T waves carry out what dimension-reduction treatment obtained;
T wave end point determination modules obtain institute for electrocardiosignal to be detected to be inputted the T waves end point determination model
State the T wave terminal point coordinates of electrocardiosignal to be detected.
A kind of computer equipment, including memory, processor and storage can be run on a memory and on a processor
Computer program, the processor realize following steps when executing described program:
Obtain T wave end point determination models;The T waves end point determination model is based on MLP neural network models to history T waves
Discrete data carries out obtained from learning training;Wherein, the history T wave discrete datas are to the history electrocardio from QT databases
T waves that signal extraction goes out carry out what dimension-reduction treatment obtained;
Electrocardiosignal to be detected is inputted into the T waves end point determination model, obtains the T of the electrocardiosignal to be detected
Wave terminal point coordinate.
A kind of computer storage media, is stored thereon with computer program, is realized when which is executed by processor following
Step:
Obtain T wave end point determination models;The T waves end point determination model is based on MLP neural network models to history T waves
Discrete data carries out obtained from learning training;Wherein, the history T wave discrete datas are to the history electrocardio from QT databases
T waves that signal extraction goes out carry out what dimension-reduction treatment obtained;
Electrocardiosignal to be detected is inputted into the T waves end point determination model, obtains the T of the electrocardiosignal to be detected
Wave terminal point coordinate.
A technical solution in above-mentioned technical proposal has the following advantages that and advantageous effect:
Above-mentioned ECG T wave end-point detection method and device obtain T waves first when to ECG T wave end point determination
The process of establishing of end point determination model, wherein T waves end point determination pattern is to carry out T waves to the history electrocardiosignal in QT databases
Extraction, and dimension-reduction treatment is carried out to the T waves after extraction and obtains history T wave discrete datas, using MLP network modes to history
T wave discrete datas carry out learning training to obtain T wave end point determination models, and electrocardiosignal to be detected is finally inputted T waves
In end point determination model, so that it may to obtain T wave terminal point coordinates.T wave ends are being obtained in above-mentioned ECG T wave end-point detection method
When point detection pattern, effectively image dimension reduction method is combined with MPL neural network models so that obtained T waves detection
Model calculation amount is smaller, and when identifying T wave terminals using T wave detection models, calculation amount is small, substantially increases the efficiency of calculating.
Description of the drawings
Fig. 1 is the flow diagram of the ECG T wave end-point detection method of the present invention in one of the embodiments;
Fig. 2 is the flow diagram of the ECG T wave end-point detection method of the present invention in one of the embodiments;
Fig. 3 is the structural schematic diagram of a wave group in electrocardiogram;
Fig. 4 is the flow diagram of the ECG T wave end-point detection method of the present invention in one of the embodiments;
Fig. 5 is the flow diagram of the ECG T wave end-point detection method of the present invention in one of the embodiments;
Fig. 6 is the flow diagram of the ECG T wave end point determination device of the present invention in one of the embodiments;
Fig. 7 is the flow diagram of the ECG T wave end point determination device of the present invention in one of the embodiments;
Fig. 8 is the structural schematic diagram of the computer equipment of the present invention in one embodiment.
Specific implementation mode
Present disclosure is described in further detail below in conjunction with preferred embodiment and attached drawing.Obviously, hereafter institute
The embodiment of description is only used for explaining the present invention rather than limitation of the invention.Based on the embodiments of the present invention, this field is general
The every other embodiment that logical technical staff is obtained without making creative work belongs to what the present invention protected
Range.It should be noted that for ease of description, only some but not all contents related to the present invention are shown in the drawings.
Fig. 1 is the flow diagram of the ECG T wave end-point detection method of the present invention in one embodiment, such as Fig. 1 institutes
Show, the ECG T wave end-point detection method in the embodiment of the present invention includes the following steps:
Step S110 obtains T wave end point determination models;T wave end point determination models be based on MLP neural network models, it is right
History T wave discrete datas carry out obtained from learning training;Wherein, history T waves discrete data is to the history from QT databases
T waves that electrocardiosignal extracts carry out what dimension-reduction treatment obtained.Electrocardiogram refers to heart in each cardiac cycle, by
Fight point, atrium, ventricle is in succession excited, along with the bioelectric variation of electrocardiogram, is drawn from body surface by electrocardiograph a variety of
The figure of the potential change of form.Common electrocardiogram is made of a series of wave group, and each wave group represents each heart
The dynamic period.
One wave group of electrocardiogram includes that P waves, QRS complex, T involve U waves.P waves are each waves produced by Atrial depolarization
First wave in group, it reflects the process of depolarization of left and right atrium.QRS complex:Including three closely coupled waves, first
Downward wave is known as Q waves, and the upright wave of the high point of after Q waves is known as R waves, and the wave after R waves under is known as S waves, QRS complex
Reflect left and right biventricular process of depolarization.T waves:T waves are located at after S-T segment, are that one relatively low and while accounting for longer wave,
It is caused by ventricular bipolar.U waves:U waves are located at after T waves, relatively low small.It can be seen that T waves are one and ventricular function
Close related waveform collection of illustrative plates, can be used for diagnose ventricle relevant disease.In addition, the phase reflects ventricle and removes between QT in electrocardiogram
The process of multipole, when patient's ventricular arrhythmia, phase duration will change between QT, therefore phase duration is for diagnosing between QT
Ventricular arrhythmia important criteria and foundation, phase duration first has to extraction QT electrocardiosignals between determining QT, however is extracting
When QT electrocardiosignals, determine that T waves final position is very crucial, and existing T waves end-point detection method type it is more (such as by
Area, template matches and statistical-simulation spectrometry etc.), but all to there is calculation amount very big for these methods, it is difficult to it is examined in real time
It surveys.
MLP (Multi-layer Perceptron, multilayer perceptron) neural network is a kind of network mould of multilayer feedforward
Type, MLP neural networks include mainly three parts:One group of perception unit composition input layer, one or more layers calculate node are hidden
The output layer of layer and one layer of calculate node.Wherein each layer all has one or more nodes, in addition to input node, each node
All it is a neuron (or processing unit) for carrying nonlinear activation function, the supervision of generally use back-propagation algorithm
Learning method is often used to train MLP, i.e. MLP neural networks are a kind of multilayer feedforward god trained according to error backpropagation algorithm
Through network, study can be carried out by the sample to a large amount of " input-output " patterns and establish mapping relations, and use steepest descent method
This learning rules is constantly adjusted the connection weight and threshold value of network by the backpropagation of the extensive error of experience, makes god
The error sum of squares approached data sample through network model is minimum.MLP can be used for that nonlinear data is identified.
In the present embodiment, learning training is carried out to history T wave discrete datas using MLP neural networks, to obtain T
Wave end point determination model.Wherein T waves end point determination pattern is that advance training is completed, need to first point signal to be detected into
When row T wave terminal point coordinates detect, only this model need to be extracted or called.
Electrocardiosignal to be detected is inputted T wave end point determination models, obtains electrocardiosignal to be detected by step S120
T wave terminal point coordinates.
Specifically, when carrying out T wave end point determinations to electrocardiosignal to be detected, electrocardiosignal to be detected is input to
In T wave end point determination models, to be assured that T waves terminal point coordinate in electrocardiosignal to be detected (i.e. sampling time point).
Above-mentioned ECG T wave end-point detection method obtains the inspection of T wave terminals first when to ECG T wave end point determination
Model is surveyed, the process of establishing of wherein T waves end point determination pattern is to carry out T wave extractions to the history electrocardiosignal in QT databases,
And to after extraction T waves carry out dimension-reduction treatment obtain history T wave discrete datas, using MLP network modes to history T waves from
It dissipates data and carries out learning training to obtain T wave end point determination models, finally by electrocardiosignal input T wave terminals inspection to be detected
It surveys in model, so that it may to obtain T wave terminal point coordinates.T wave end point determinations are being obtained in above-mentioned ECG T wave end-point detection method
When pattern, effectively image dimension reduction method is combined with MPL neural network models so that obtained T wave detection model meters
Calculation amount is smaller, and when identifying T wave terminals using T wave detection models, calculation amount is small, substantially increases the efficiency of calculating.
In one of the embodiments, as shown in Fig. 2, the T waves discrete data include the discrete sampling time and with it is discrete
Sampling time corresponding target output value;T wave end point determination patterns can be obtained by following steps;
Step S130 obtains history electrocardiosignal from QT databases.
Step S140 carries out signal subsection processing to history electrocardiosignal using the first preset window function, obtains each T waves
Electrocardiosignal.
Step S150 carries out dimension-reduction treatment to each T waves electrocardiosignal, obtains each destination sample time and adopted with each target
Sample time corresponding target output voltage value.
Step S160, it is corresponding to each destination sample time and each destination sample time using MLP neural network models
Target output voltage value carries out learning training, obtains T wave end point determination models.
Specifically, before using MLP neural network models, history T wave discrete datas first to be extracted.First from QT data
Extraction history ECG signal in library (i.e. history ECG data library), a total of 105 electrocardiographic recorders of QT databases, each
Record includes electrocardiogram (ECG) data for two duration 15min that sample rate is 250Hz, wherein further includes the heart in each record
Popular name for expert claps annotation at least 30 hearts, accurate marker QRS complex, P waves, the peak point of T waves, beginning and end position
It sets and T wave morphologies.As soon as one of heart bat refers to a cardiac cycle, the bat of each heart will produce an electrocardiagraphic wave group.
The history ECG signal extracted from QT databases is the ECG signal that record has wave group, in order to improve
The rate of image procossing considers T wave frequency ranges, is carried out at segmentation to history electrocardiogram using the first preset window function
Reason, when being segmented, is segmented history electrocardiogram as starting point using R crest values point.Result such as Fig. 3 institutes after segmentation
Show, wherein characteristic is generated it is found that between the 200ms~600ms of T waves mostly after the r-wave according to electrocardiogram, then when T waves sample
Between the expression formula of (i.e. abscissa) be:
xi=S [Ri+200:Ri+600];
T wave sampling time (i.e. abscissa) is standardized, the expression formula for obtaining target output is:
Wherein S indicates electrocardiosignal, RiIndicate position (or moment, that is, when sampling of R crest values point in i-th of wave group
Between), xirefIndicate T wave start positions, Te in i-th of wave groupiIndicate the T wave terminals that expert marks in QT i-th of wave group of database
Position, by divided by 400, can be by yiIt is normalized between 0~1.After being segmented to history electrocardiosignal, it can obtain multiple
T wave electrocardiosignals.
In addition, since the sample frequency of QT databases is generally 250Hz (or the sampling time is 4ms), then in entire T waves
(i.e. in [R in electrocardiosignali+200:Ri+ 600] in range) share 100 points, i.e. xiBe one 100 dimension vector, dimension compared with
Height, when training MLP neural network models using the T wave ecg signal datas of higher-dimension, calculation amount is very big.In the present embodiment
In, to the x of T wave electrocardiosignalsiDimension-reduction treatment is carried out, obtains each discrete sampling time, and corresponding with each discrete sampling time
Target output value.Then using each discrete sampling time and target output value corresponding with each discrete sampling time as training sample
It is right, learning training is carried out to MLP neural network models, obtains T wave end point determination models.In the present embodiment, T wave electrocardios are believed
Number xiThere are many ways to carrying out dimension-reduction treatment, discrete fourier variation, discrete cosine transform, principal component analysis may be used
Method and down-sampled etc..MLP models are trained using T waves discrete data (the T wave numbers evidence i.e. after dimensionality reduction), calculating can be effectively reduced
Amount, improves model training rate.
Outside it, the first preset window function is a kind of for extracting T wave electrocardiosignals;Utilize the first preset window letter
Number processing history electrocardiosignal, you can obtain T wave electrocardiosignals.
In a kind of optional embodiment, use hidden layer for 1, node in hidden layer is 1~32, and input layer is
1~16 MLP neural network models to carry out learning training to history T wave discrete datas.It is distributed and is advised according to electrocardiogram (i.e. T waves)
Rule selects above-mentioned specific MLP neural network models, on the one hand can improve model training speed, on the other hand, complete using training
At model come when determining the wave final positions T, testing result is accurate.
Before the step of carrying out signal subsection to history electrocardiosignal in one of the embodiments,:
History electrocardiosignal is filtered.
Specifically, it during testing electrocardiogram, is sometimes prone to generate interference, occurs clutter in electrocardiogram, these
The analysis of clutter meeting ECG signal generates interference.Therefore, after obtaining history electrocardiosignal in QT databases, to the history heart
Before electric signal carries out signal subsection processing, history electrocardiosignal is filtered, removes clutter.Using filtered history
Electrocardiosignal extracts T waves discrete datas, and (i.e. each discrete sampling time and target corresponding with each discrete sampling time are defeated
Go out value) it is more accurate, so use the T waves detection model that above-mentioned T waves discrete data training MLP neural network models obtain more
Accurately, so that carrying out the detection of T waves final position, obtained detection to electrocardiosignal to be detected using T waves detection model
As a result just more accurate.
Optionally, generally use bandpass filtering is filtered history electrocardiogram (ECG) data, that is, selects specific frequency model
Interior electrocardiogram wave band is enclosed, reduces calculation amount, substantially increases computational efficiency.
Include the step of being filtered to history electrocardiosignal in one of the embodiments,:
Bandpass filtering treatment is carried out to history electrocardiosignal using Chebyshev's bandpass filter.
Specifically, T wave frequency ranges are between 0~10Hz in electrocardiogram, wherein 92% energy all concentrates on 1~8Hz
Between, history electrocardiosignal is handled using Chebyshev's bandpass filter in the present embodiment, wherein bandpass filtering
Range can be 0~50Hz.History electrocardiosignal is handled using Chebyshev's bandpass filter, can effectively obtain the useful heart
Electric signal removes the interference of some useless clutters, on the one hand reduces interference, improve accuracy rate;On the other hand, it reduces and calculates
Amount.In addition, Chebyshev filter is the filter of the ripples such as frequency response amplitude on passband or stopband, intermediate zone ratio
Relatively narrow, envelope eapsulotomy is good.
In one of the embodiments, as shown in figure 4, in the step of carrying out dimension-reduction treatment to T wave electrocardiosignals, packet
It includes:
Step S152 carries out dimension-reduction treatment using discrete cosine transform to T wave electrocardiosignals.
Specifically, DCT (Discrete Cosine Transform, discrete cosine transform) is converted, and is signal processing sum number
According to a kind of mode of compression, essence is exactly to indicate origin number using fewer number of points in the case of certain loss,
The data of a higher-dimension essence can be indicated after dct transform with fewer number of digit, and the essence of data is special
Sign does not change.In the present embodiment, dimension-reduction treatment is carried out to T wave electrocardiosignals using discrete cosine transform, i.e., to T waves
X in electrocardiosignali([Ri+200:Ri+ 600] dimension-reduction treatment) is carried out, the expression formula that target exports after dimensionality reduction is:
Wherein, target output value when y (k) representation dimensions are k, k representation dimensions, xnIndicate n-th of sampling time.
For the ease of understanding the present embodiment, a detailed embodiment is provided.A plurality of record is chosen from QT databases
History electrocardiosignal, is filtered history electrocardiosignal and segment processing, n T wave electrocardiosignal is obtained, to each
A T waves electrocardiosignal carries out dimension-reduction treatment using discrete cosine transform, obtains multiple discrete sampling time x1~xnWith with it is multiple from
Dissipate the sampling time corresponding target output value y1~yn, then by x1~xnAnd y1~ynIt forms MLP neural network models and trains sample
This obtains T wave end point determination models to neural metwork training after the completion of training.
In one of the embodiments, as shown in figure 4, further including:In the step of carrying out dimension-reduction treatment to T wave electrocardiosignals
In, further include:
Step S154 carries out dimension-reduction treatment using Principal Component Analysis to T wave electrocardiosignals.
Specifically, principal component analysis (Principal Component Analysis) is a kind of method of mathematic(al) manipulation,
One group of given correlated variables is changed into another group of incoherent variable by it by linear transformation, these new variables are according to variance
That successively decreases successively is ranked sequentially.It keeps the population variance of variable constant in mathematic(al) manipulation, makes the first variable that there is maximum variance,
Referred to as first principal component, bivariate variance time is big, and uncorrelated with the first variable, referred to as Second principal component,.Class successively
It pushes away, I variable just has I principal component.Principal Component Analysis is exactly to be gone to explain the major part in original data with less variable
Variable, by the very high variables transformations of many correlations in our hands at being mutually independent or incoherent variable.Typically select
Go out fewer than original variable number, can explain several new variables of variable in most of data, i.e., so-called principal component, principal component analysis
Actually a kind of dimension reduction method, i.e., sample is described with a kind of small number of feature reduces feature space dimension to reach
Several methods, this method can eliminate the relative influence between evaluation index, and calculation amount is smaller and specification.
In one of the embodiments, as shown in figure 4, in the step of carrying out dimension-reduction treatment to T wave electrocardiosignals, also wrap
It includes:
Step S156 carries out dimension-reduction treatment using down-sampled method to T wave electrocardiosignals.
Specifically, it is down-sampled be reduce signal specific sample rate process, commonly used in reduce message transmission rate or
Person's size of data is a kind of very simple method for quickly carrying out dimensionality reduction to data, and calculating process is simple, easy to operate.
In one of the embodiments, as shown in figure 5, electrocardiosignal to be detected is inputted the T waves end point determination
Model further includes in the step of obtaining the T wave terminal point coordinates of the electrocardiosignal to be detected:
Step S122 is filtered electrocardiosignal to be detected;
Step S124 carries out signal point using the second preset window function to the electrocardiosignal to be detected after being filtered
Section processing, obtains T wave electrocardiosignals to be detected;
T wave electrocardiosignals to be detected are input to T wave end point determination models, obtain the T wave hearts to be detected by step S126
The T wave terminal point coordinates of electric signal.
Specifically, when carrying out T wave end point determinations to electrocardiosignal to be detected, first electrocardiosignal to be detected is carried out
It is filtered and signal subsection is handled, obtain T wave electrocardiosignals to be detected, it then will be defeated to the T wave electrocardiosignals of detection
Enter the T wave terminal point coordinates that T wave electrocardiosignals to be detected are obtained to T wave end point determination models.In the present embodiment, it first treats
The electrocardiosignal of detection is filtered, then using the second preset window function to the electrocardio to be detected after being filtered
Signal carries out the segment processing of signal, obtains T wave electrocardiosignals to be detected, and it is miscellaneous to remove some first put in signal to be detected
Wave reduces interference, convenient for quickly and accurately determining T wave terminal point coordinates in electrocardiosignal to be detected.
According to the ECG T wave end-point detection method of aforementioned present invention, the present invention also provides a kind of inspections of ECG T wave terminal
Device is surveyed, below in conjunction with the accompanying drawings and the ECG T wave end point determination device of the present invention is described in detail in preferred embodiment.
Fig. 6 is the structural schematic diagram of the ECG T wave end point determination device of the present invention in one embodiment.Such as Fig. 6 institutes
Show, the ECG T wave end point determination device in the embodiment, including:
Model acquisition module 10, for obtaining T wave end point determination models;The T waves end point determination model is to be in advance based on
MLP neural network models, obtained from carrying out learning training to history T wave discrete datas, wherein the history T wave discrete datas
Be in QT databases history electrocardiosignal carry out T wave extractions, and to after extraction the T waves carry out dimension-reduction treatment obtain and
It obtains;
T wave end point determinations module 20 is obtained for electrocardiosignal to be detected to be inputted the T waves end point determination model
The T wave terminal point coordinates of the electrocardiosignal to be detected.
In one of the embodiments, as shown in fig. 7, ECG T wave end point determination device, further includes:
Electrocardiosignal acquisition module 30, for obtaining history electrocardiosignal from the QT databases;
T waves obtain module 40, for carrying out signal subsection to the history electrocardiosignal using the first preset window function
Processing, obtains T wave electrocardiosignals;
Centrifugal pump obtains module 50, for carrying out dimension-reduction treatment to the T waves electrocardiosignal, obtains each discrete sampling time
And target output value corresponding with each discrete sampling time;
Detection model obtain module 60, for use the MLP neural network models, to each discrete sampling time with
And state corresponding target output value of each discrete sampling time and carry out learning training, obtain the T waves end point determination model.
Further include in one of the embodiments,:
Filter module, for being filtered to the history electrocardiosignal.
Further include in one of the embodiments,:
The filter module is additionally operable to carry out band logical filter to the history electrocardiosignal using Chebyshev's bandpass filter
Wave processing.
Centrifugal pump obtains module 50 and is additionally operable to using discrete cosine transform to the T waves heart in one of the embodiments,
Electric signal carries out dimension-reduction treatment.
Centrifugal pump is obtained module 50 and is additionally operable to believe T wave electrocardios using Principal Component Analysis in one of the embodiments,
Number carry out dimension-reduction treatment.
Centrifugal pump obtains module 50 and is additionally operable to using down-sampled method to T wave electrocardiosignals in one of the embodiments,
Carry out dimension-reduction treatment.
Further include in one of the embodiments,:
The filter module is additionally operable to be filtered the electrocardiosignal to be detected;
The T waves obtain module, are additionally operable to using the second preset window function to described to be detected after being filtered
Electrocardiosignal carries out signal subsection, obtains T wave electrocardiosignals to be detected;
T wave end point determination modules, are additionally operable to state the T wave electrocardiosignals to be detected are input to the T waves terminal inspection
Model is surveyed, the T wave terminal point coordinates of the T wave electrocardiosignals to be detected are obtained.
Above-mentioned ECG T wave end point determination device can perform the ECG T wave end point determination that the embodiment of the present invention is provided
Method has the corresponding function module of execution method and advantageous effect.As for the processing method performed by wherein each function module,
Such as model acquisition module 10, T wave end point determinations module 20, centrifugal pump obtain module 50, can refer in above method embodiment
Description, no longer repeated herein.
According to the ECG T wave end-point detection method and device of aforementioned present invention, the present invention also provides a kind of computers to set
Standby, below in conjunction with the accompanying drawings and the computer equipment of the present invention is described in detail in preferred embodiment.
Fig. 8 is the structural schematic diagram of the computer equipment of the present invention in one embodiment.As shown in figure 8, the embodiment
In computer equipment 800, including memory 802, processor 804 and storage can run on a memory and on a processor
Computer program, wherein processor can realize all method and steps in the method for the present invention embodiment when executing program.
Processor 804 can perform the ECG T wave terminal inspection that the embodiment of the present invention is provided in above computer equipment 800
Survey method has the corresponding advantageous effect of execution method.The description in above method embodiment is can refer to, is no longer gone to live in the household of one's in-laws on getting married herein
It states.
According to ECG T wave end-point detection method, device and the computer equipment of aforementioned present invention, the present invention also provides one
Kind computer readable storage medium, below in conjunction with the accompanying drawings and preferred embodiment carries out the computer readable storage medium of the present invention
It is described in detail.
Computer readable storage medium in the embodiment of the present invention, is stored thereon with computer program, which is handled
All method and steps in the method for the present invention embodiment may be implemented in device when executing.
One of ordinary skill in the art will appreciate that realizing all or part of flow in above-described embodiment method, being can be with
Relevant hardware is instructed to complete by computer program.Computer program can be stored in a computer read/write memory medium
In, the program is when being executed, it may include such as the flow of the embodiment of above-mentioned each method.Wherein, storage medium can be magnetic disc, light
Disk, read-only memory (Read-Only Memory, ROM) or random access memory (Random Access Memory,
RAM) etc. ".
Above computer readable storage medium storing program for executing is for storing the ECG T wave end point determination side that the embodiment of the present invention is provided
The program (instruction) of method, wherein the ECG T wave end point determination side that the embodiment of the present invention is provided can be executed by executing the program
Method has the corresponding advantageous effect of execution method.The description in above method embodiment is can refer to, is no longer repeated herein.
Each technical characteristic of above example can be combined arbitrarily, to keep description succinct, not to above-described embodiment
In each technical characteristic it is all possible combination be all described, as long as however, the combination of these technical characteristics be not present lance
Shield is all considered to be the range of this specification record.
Only several embodiments of the present invention are expressed for above example, the description thereof is more specific and detailed, but can not
Therefore it is construed as limiting the scope of the patent.It should be pointed out that for those of ordinary skill in the art,
Under the premise of not departing from present inventive concept, various modifications and improvements can be made, these are all within the scope of protection of the present invention.
Therefore, the protection domain of patent of the present invention should be determined by the appended claims.
Claims (11)
1. a kind of ECG T wave end-point detection method, which is characterized in that include the following steps:
Obtain T wave end point determination models;The T waves end point determination model is discrete to history T waves based on MLP neural network models
Data carry out obtained from learning training;Wherein, the history T wave discrete datas are to the history electrocardiosignal from QT databases
The T waves that extract carry out what dimension-reduction treatment obtained;
Electrocardiosignal to be detected is inputted into the T waves end point determination model, the T waves for obtaining the electrocardiosignal to be detected are whole
Point coordinates.
2. ECG T wave end-point detection method according to claim 1, which is characterized in that the T waves discrete data includes
Discrete sampling time and target output value corresponding with the discrete sampling time;The T waves end point determination model passes through following
Step obtains:
The history electrocardiosignal is obtained from the QT databases;
Signal subsection processing is carried out to the history electrocardiosignal using the first preset window function, obtains each T waves electrocardiosignal;
Dimension-reduction treatment is carried out to each T waves electrocardiosignal, obtains each discrete sampling time and when with each discrete sampling
Between corresponding target output value;
It is corresponding to each discrete sampling time and each discrete sampling time using the MLP neural network models
Target output value carries out learning training, obtains the T waves end point determination model.
3. ECG T wave end-point detection method according to claim 2, which is characterized in that the history electrocardiosignal
Before the step of carrying out signal subsection, further include
The history electrocardiosignal is filtered.
4. ECG T wave end-point detection method according to claim 3, which is characterized in that believe to the history electrocardio
The step of number being filtered include:
Bandpass filtering treatment is carried out to the history electrocardiosignal using Chebyshev's bandpass filter.
5. according to claim 2-4 any one of them ECG T wave end-point detection methods, which is characterized in that the T waves
Electrocardiosignal carried out in the step of dimension-reduction treatment, including:
Dimension-reduction treatment is carried out to the T waves electrocardiosignal using discrete cosine transform.
6. according to claim 2-4 any one of them ECG T wave end-point detection methods, which is characterized in that the T waves
Electrocardiosignal carried out in the step of dimension-reduction treatment, further included:
Dimension-reduction treatment is carried out to the T waves electrocardiosignal using Principal Component Analysis.
7. according to claim 2-4 any one of them ECG T wave end-point detection methods, which is characterized in that the T waves
Electrocardiosignal carried out in the step of dimension-reduction treatment, further included:
Dimension-reduction treatment is carried out to the T waves electrocardiosignal using down-sampled method.
8. according to claim 1-4 any one of them ECG T wave end-point detection methods, which is characterized in that will be to be detected
Electrocardiosignal the step of inputting the T waves end point determination model, obtaining the T wave terminal point coordinates of the electrocardiosignal to be detected
In further include:
The electrocardiosignal to be detected is filtered;
Signal subsection processing is carried out to the electrocardiosignal to be detected after being filtered using the second preset window function, is obtained
To T wave electrocardiosignals to be detected;
The T wave electrocardiosignals to be detected are input to the T waves end point determination model, obtain the T wave hearts to be detected
The T wave terminal point coordinates of electric signal.
9. a kind of ECG T wave end point determination device, which is characterized in that including:
Model acquisition module, for obtaining T wave end point determination models;The T waves end point determination model is based on MLP neural networks
Model carries out history T wave discrete datas obtained from learning training;Wherein, the history T wave discrete datas are to from QT numbers
The T waves that are extracted according to history electrocardiosignal in library carry out what dimension-reduction treatment obtained;
T wave end point determination modules obtain described wait for for electrocardiosignal to be detected to be inputted the T waves end point determination model
The T wave terminal point coordinates of the electrocardiosignal of detection.
10. a kind of computer equipment, including memory, processor and storage are on a memory and the meter that can run on a processor
Calculation machine program, which is characterized in that the step of processor realizes claim 1-8 the methods when executing described program.
11. a kind of computer storage media, is stored thereon with computer program, which is characterized in that the program is executed by processor
The step of Shi Shixian claim 1-8 the methods.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810214428.4A CN108309284A (en) | 2018-03-15 | 2018-03-15 | ECG T wave end-point detection method and device |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810214428.4A CN108309284A (en) | 2018-03-15 | 2018-03-15 | ECG T wave end-point detection method and device |
Publications (1)
Publication Number | Publication Date |
---|---|
CN108309284A true CN108309284A (en) | 2018-07-24 |
Family
ID=62902218
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810214428.4A Pending CN108309284A (en) | 2018-03-15 | 2018-03-15 | ECG T wave end-point detection method and device |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108309284A (en) |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109745037A (en) * | 2019-02-14 | 2019-05-14 | 广州视源电子科技股份有限公司 | A kind of image processing method, device, equipment and storage medium |
CN110897627A (en) * | 2018-09-14 | 2020-03-24 | 杭州脉流科技有限公司 | Electrocardiogram signal feature extraction method, device, equipment, system and storage medium |
CN111803061A (en) * | 2020-06-28 | 2020-10-23 | 武汉中旗生物医疗电子有限公司 | R wave identification method and device based on target detection |
CN111956203A (en) * | 2020-08-18 | 2020-11-20 | 广州视源电子科技股份有限公司 | Electrocardiosignal parameterization method, model training method, device, equipment and medium |
CN112587152A (en) * | 2020-11-11 | 2021-04-02 | 上海数创医疗科技有限公司 | 12-lead T wave extraction method fusing U-net network and filtering method |
CN112587151A (en) * | 2020-11-11 | 2021-04-02 | 上海数创医疗科技有限公司 | 12-lead T wave extraction device fusing U-net network and filtering method |
CN115770048A (en) * | 2021-09-08 | 2023-03-10 | 上海联影医疗科技股份有限公司 | Electrocardiosignal processing method and device, computer equipment and storage medium |
CN117076322A (en) * | 2023-08-30 | 2023-11-17 | 合芯科技(苏州)有限公司 | Method and system for detecting abnormal working mode of research and development technology service provider |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO1993018473A1 (en) * | 1992-03-10 | 1993-09-16 | Siemens Aktiengesellschaft | Preclassification process of high-dimensional characteristic vectors for signal processing purposes |
CN101919695A (en) * | 2010-08-06 | 2010-12-22 | 李楚雅 | Electrocardiosignal QRS complex detection method based on wavelet transform |
CN104523264A (en) * | 2014-12-31 | 2015-04-22 | 深圳职业技术学院 | Electrocardiosignal processing method |
CN106073755A (en) * | 2016-05-27 | 2016-11-09 | 成都信汇聚源科技有限公司 | The implementation method that in a kind of miniature holter devices, atrial fibrillation identifies automatically |
-
2018
- 2018-03-15 CN CN201810214428.4A patent/CN108309284A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO1993018473A1 (en) * | 1992-03-10 | 1993-09-16 | Siemens Aktiengesellschaft | Preclassification process of high-dimensional characteristic vectors for signal processing purposes |
CN101919695A (en) * | 2010-08-06 | 2010-12-22 | 李楚雅 | Electrocardiosignal QRS complex detection method based on wavelet transform |
CN104523264A (en) * | 2014-12-31 | 2015-04-22 | 深圳职业技术学院 | Electrocardiosignal processing method |
CN106073755A (en) * | 2016-05-27 | 2016-11-09 | 成都信汇聚源科技有限公司 | The implementation method that in a kind of miniature holter devices, atrial fibrillation identifies automatically |
Non-Patent Citations (1)
Title |
---|
LEON A A S , MOLINA D M , SEISDEDOS C R V , ET AL: "Neural Network Approach for T -wave End Detection: A Comparison of Architectures", 《COMPUTING IN CARDIOLOGY CONFERENCE (CINC). IEEE》 * |
Cited By (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110897627A (en) * | 2018-09-14 | 2020-03-24 | 杭州脉流科技有限公司 | Electrocardiogram signal feature extraction method, device, equipment, system and storage medium |
CN110897627B (en) * | 2018-09-14 | 2023-04-18 | 杭州脉流科技有限公司 | Electrocardiogram signal feature extraction method, device, equipment, system and storage medium |
CN109745037A (en) * | 2019-02-14 | 2019-05-14 | 广州视源电子科技股份有限公司 | A kind of image processing method, device, equipment and storage medium |
CN111803061A (en) * | 2020-06-28 | 2020-10-23 | 武汉中旗生物医疗电子有限公司 | R wave identification method and device based on target detection |
CN111803061B (en) * | 2020-06-28 | 2022-12-30 | 武汉中旗生物医疗电子有限公司 | R wave identification method and device based on target detection |
CN111956203A (en) * | 2020-08-18 | 2020-11-20 | 广州视源电子科技股份有限公司 | Electrocardiosignal parameterization method, model training method, device, equipment and medium |
CN112587152A (en) * | 2020-11-11 | 2021-04-02 | 上海数创医疗科技有限公司 | 12-lead T wave extraction method fusing U-net network and filtering method |
CN112587151A (en) * | 2020-11-11 | 2021-04-02 | 上海数创医疗科技有限公司 | 12-lead T wave extraction device fusing U-net network and filtering method |
CN112587152B (en) * | 2020-11-11 | 2022-08-16 | 上海数创医疗科技有限公司 | 12-lead T wave extraction method fusing U-net network and filtering method |
CN115770048A (en) * | 2021-09-08 | 2023-03-10 | 上海联影医疗科技股份有限公司 | Electrocardiosignal processing method and device, computer equipment and storage medium |
CN117076322A (en) * | 2023-08-30 | 2023-11-17 | 合芯科技(苏州)有限公司 | Method and system for detecting abnormal working mode of research and development technology service provider |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108309284A (en) | ECG T wave end-point detection method and device | |
CN108714026B (en) | Fine-grained electrocardiosignal classification method based on deep convolutional neural network and online decision fusion | |
CN106108889B (en) | Electrocardiogram classification method based on deep learning algorithm | |
Cao et al. | Atrial fibrillation detection using an improved multi-scale decomposition enhanced residual convolutional neural network | |
CN109222963A (en) | A kind of anomalous ecg method for identifying and classifying based on convolutional neural networks | |
CN110313894A (en) | Arrhythmia cordis sorting algorithm based on convolutional neural networks | |
Farag | A self-contained STFT CNN for ECG classification and arrhythmia detection at the edge | |
Kshirsagar et al. | Classification of ECG-signals using artificial neural networks | |
Rohmantri et al. | Arrhythmia classification using 2D convolutional neural network | |
JP7487965B2 (en) | Prediction method of electrocardiogram heart rate multi-type based on graph convolution | |
Kumari et al. | Performance evaluation of neural networks and adaptive neuro fuzzy inference system for classification of Cardiac Arrhythmia | |
Barman et al. | Rough set based segmentation and classification model for ECG | |
Sherly et al. | An efficient honey badger based Faster region CNN for chronc heart Failure prediction | |
Shukri et al. | Investigation on Elman neural network for detection of cardiomyopathy | |
Kishore et al. | An innovative machine learning approach for classifying ECG signals in healthcare devices | |
Zubair et al. | Removal of motion artifacts from ECG signals by combination of recurrent neural networks and deep neural networks | |
Jiang et al. | Heartbeat classification system based on modified stacked denoising autoencoders and neural networks | |
Shaker et al. | Heartbeat classification using 1D convolutional neural networks | |
CN112022149B (en) | Atrial fibrillation detection method based on electrocardiosignals | |
Khandait et al. | ECG signal processing using classifier to analyses cardiovascular disease | |
Nanarkar et al. | A Survey on Classification and identification of Arrhythmia using Machine Learning techniques | |
Maji et al. | Detection and characterisation of QRS complex in VMD domain | |
Moein et al. | Intelligent ECG signal noise removal using PSONN | |
CN116570295B (en) | Electrocardiogram low-voltage T-wave end point positioning method | |
Srinivasulu et al. | Novel method to find the parameter for noise removal from multi-channel ecg waveforms |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180724 |
|
RJ01 | Rejection of invention patent application after publication |