CN107837083B - J wave automatic testing method based on least square method supporting vector machine - Google Patents
J wave automatic testing method based on least square method supporting vector machine Download PDFInfo
- Publication number
- CN107837083B CN107837083B CN201711037915.XA CN201711037915A CN107837083B CN 107837083 B CN107837083 B CN 107837083B CN 201711037915 A CN201711037915 A CN 201711037915A CN 107837083 B CN107837083 B CN 107837083B
- Authority
- CN
- China
- Prior art keywords
- wave
- electrocardiosignal
- vector machine
- square method
- supporting vector
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
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
Landscapes
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Cardiology (AREA)
- Heart & Thoracic Surgery (AREA)
- Molecular Biology (AREA)
- Pathology (AREA)
- Engineering & Computer Science (AREA)
- Biomedical Technology (AREA)
- Physics & Mathematics (AREA)
- Medical Informatics (AREA)
- Biophysics (AREA)
- Surgery (AREA)
- Animal Behavior & Ethology (AREA)
- General Health & Medical Sciences (AREA)
- Public Health (AREA)
- Veterinary Medicine (AREA)
- Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)
Abstract
The present invention relates to the detection of J wave identification and classification, specially the J wave automatic testing method based on least square method supporting vector machine.It since J wave is often blended in ST sections, often shows as ST sections and raises, mainly include hump type, three kinds of frustration form, spike type Main Morphologies.When the amplitude very little of J wave, it still seems to have no difference after mixing with normal electrocardiosignal with normal electrocardiosignal, it is therefore necessary to extract it and be analyzed and sorted out, facilitate clinic and provide alert.It relies only on doctor's clinical experience and the requirement that diagnosis is extremely difficult to high accuracy is carried out to J wave.This patent overcomes low to the accuracy in detection of existing J wave and classification method insufficient defect, feature extraction is carried out to J wave from signal processing angle, and widely used machine learning method classifies to extracted feature in two classification problems at present for combination, is detected automatically using computer to J wave to realize.
Description
Technical field
The present invention relates to the detection of J wave identification and classification, and specially the J wave based on least square method supporting vector machine is examined automatically
Survey method.
Background technique
Cardiovascular disease is very big to human life and health harm, and in China, heart disease rate increases always, its serious danger
Do harm to the physical and mental health of people.Currently, the whole world has nearly 17,000,000 people to die of cardiovascular disease every year, death toll is accounted for entirely
The nearly one third of all death tolls of ball.In recent years, J wave as electrocardiogram (ECG) ventricular bipolar New Set increasingly by
To clinical attention, become the important reference indicator that doctor judges conditions of patients in the clinical diagnosis of cardiovascular disease.
Medically usually QRS wave end of cluster and ST interphase start between tie point call J point, it represents cardiac ventricles
The end of depolarization and the beginning of multipole.When J point or ST sections of form, time limit and amplitude significantly change, show at least
J point raises 0.1m V or more from base line shifts on two consecutive leads, and the duration reaches 20ms (standard is not unique at present),
It combines together to form the waveform of dome-shaped, hump shape, needle pattern with the ascending branch of T wave, referred to as J wave (as shown in Figure 1).When
J popin is moved or is raised, then can imply the appearance of cardiac event (such as ventricle overruns, ventricular fibrillation or coronary heart disease), or even is led
Fatal arrhythmia cordis, or even the generation of sudden death.If finding J wave in electrocardiogram, which is likely to generate the heart
Dirty event.If doctor can timely have found J wave, he can understand patient's state of an illness as soon as possible, make diagnosis.Institute
To improve the distinguishing ability to normal ECG variation J wave and abnormal J wave, facilitate the high-risk trouble for identifying clinical abnormal J wave
Person reduces the generation of malignant arrhythmia and the sudden death of idiopathic ventricular fibrillation, has very big clinical meaning and realistic meaning.
But currently, doctor mainly removes detection J from electrocardiogram, monocardiogram by the naked eyes of oneself and clinical experience
Wave is difficult accurately to detect.So the present invention utilizes meter in conjunction with current state-of-the-art machine learning method from signal processing angle
Calculation machine detects J wave.
Document An Analytic Wavelet Transform with a Flexible Time-Frequency
Covering discloses flexible analytical wavelet, and specifically disclosing flexibly parsing small echo is realized by one group of iterative filter group
, it mainly include a low-pass filter and two high-pass filters, one of high-pass filter is analysis positive frequency, other
Two filters are used to analyze negative frequency, so as to the ginseng such as neatly Control platform factor Q, redundancy R and scale factor d
Number.Brain electrosleep of the document based on fuzzy entropy describes fuzzy entropy in feature extraction and classifying by stages, and fuzzy entropy is with an index letter
Digital-to-analogue is gelatinized similarity measurement formula, overcomes the limitation that Sample Entropy defines, and enables fuzzy entropy with Parameters variation transition
Smoothly, and have the relative uniformity and short data collection treatment characteristic of Sample Entropy.The present invention using flexible analytical wavelet and
The characteristic of fuzzy entropy come improve J wave detection accuracy.
Summary of the invention
It since J wave is often blended in ST sections, often shows as ST sections and raises, mainly include hump type, frustration form, spike type three
Kind Main Morphology.When the amplitude very little of J wave, it still seems with normal electrocardiosignal after mixing with normal electrocardiosignal
It has no difference, it is therefore necessary to it be extracted and analyzed and sorted out, facilitate clinic and provide alert.The present invention in order to
Solve the problems, such as that relying only on doctor's clinical experience carries out the requirement that diagnosis is extremely difficult to high accuracy to J wave, provides based on most
Small two multiply the J wave automatic testing method of support vector machines.
The present invention adopts the following technical scheme that realization: the J wave based on least square method supporting vector machine detects automatically
Method, comprising the following steps:
(1) ecg signal data collection is divided into training set, verifying collection and test set, training set to verify in collection and test set
All include normal electrocardiosignal and the wave containing J electrocardiosignal, ecg signal data collection is pre-processed, i.e., to electrocardiosignal into
Row denoising goes baseline drift, removes artefact etc., and carries out the heart to pretreated electrocardiosignal and clap segment processing;
(2) five layers of flexibly parsing wavelet decomposition then are carried out to the electrocardiosignal after segmentation, taken a message to obtain five stratons
Number (decompose detail coefficients);
(3) fuzzy entropy of this each layer of five stratons band signal is calculated, is then believed using this five layers fuzzy entropy as electrocardio
Number feature vector;
(4) using the feature vector of ecg signal data concentration training collection as least square method supporting vector machine (LS-SVM) point
The input of class device, is trained least square method supporting vector machine, and carries out model selection by the feature vector of verifying collection, most
Test set data are tested afterwards, obtain final mask.
(5) electrocardiosignal feature vector to be measured is input in trained least square method supporting vector machine final mask,
To realize the Classification and Identification of normal electrocardiosignal Yu the electrocardiosignal of wave containing J, J wave signal is successfully detected.
This patent overcomes low to the accuracy in detection of existing J wave and classification method insufficient defect, from signal processing angle
Degree, which sets out, carries out feature extraction to J wave, and combine at present in two classification problems widely used machine learning method to being mentioned
It takes feature to classify, J wave is detected automatically using computer to realize.
Detailed description of the invention
Fig. 1 is the electro-cardiologic signal waveforms figure of the wave containing J.
Fig. 2 is flow chart of the present invention.
Specific embodiment
It mainly include three in the J wave automatic testing method based on least square method supporting vector machine designed in the technical program
A part.First part is that flexibly parsing wavelet decomposition is carried out to electrocardiosignal (ECG signal), and wavelet transformation is in the latest 20 years
In the widely used technology that field of signal processing is risen, with development is continuously improved, flexible wavelet transformation is herein
On the basis of improved fresh approach, this method has good effect in terms of biomedicine signals.
Second part is the calculating to hierarchical subband signal ambiguity entropy, and fuzzy entropy is similar between a kind of two sequences of measurement
The method of property, is the improvement of Sample Entropy, while inheriting the relative uniformity and short data collection treatment characteristic of Sample Entropy, fuzzy entropy
It can be effective
Part III is to be classified using the least square method supporting vector machine in machine learning method to feature is extracted, most
Small two multiply the improved method that support vector machines is support vector machines, overcome hyperplane parameter present in algorithm of support vector machine
Matrix size is influenced very big by training number of samples in selection and quadratic problem solution, causes solution scale is excessive to ask
Topic.
First part, the ECG signal based on flexible analytical wavelet are decomposed
Flexible analytical wavelet has the good characteristics such as translation invariance, the covering of flexible time-frequency and adjustable oscillation base.
Its method relative to other improvements wavelet transformation, due to having the wavelet basis of a pair of of Hilbert transform pairs, so as to spirit
Ground living Control platform factor Q, redundancy R and scale factor d etc. parameters.Flexibly parsing small echo is by one group of iterative filter
What group was realized, mainly include a low-pass filter and two high-pass filters, and one of high-pass filter is the positive frequency of analysis
Rate, other two filters are used to analyze negative frequency.
The frequency response expression formula of low-pass filter are as follows:
Wherein, p and q respectively represent low pass
The up-sampling parameter and down-sampling parameter of filter, ωsAnd ωpIt indicates the stopband and band connection frequency of low-pass filter, distinguishes table
Show as follows:
The response frequency expression formula of high-pass filter are as follows:
Wherein, r and s generation respectively
The up-sampling parameter and down-sampling parameter of table high-pass filter.Parameter ω0,ω1,ω2,ω3, it respectively indicates as follows:
In this algorithm, intermediate zone selection are as follows:for w∈[0,
π], all parameters of flexible analytical wavelet meet following relationship: When intermediate zone meets following equation, signal can perfect reconstruction: | θ (w) |2+|θ(π-w)
|2=1,
Second part, the feature extraction based on fuzzy entropy:
As the improvement of Sample Entropy (Samp En) algorithm, fuzzy entropy is blurred similarity measurement formula with an exponential function,
Enabling fuzzy entropy with Parameters variation transitions smooth, and in the case where parameter value very little, its definition is still significant, together
When inherit the relative uniformity and short data collection treatment characteristic of Sample Entropy.It is calculated shown in steps are as follows:
(1) N point sampling sequence is set as { u (i): 1≤i≤N };
(2) reconstruct generates one group of m n dimensional vector n in order of sequenceWherein u (i), u (i+1) ...,
U (i+m-1) } represent the value of continuous m u since i-th point, u0It (i) is its mean value,
(3) vector is definedWithBetween distanceIt is difference in the two corresponding element maximum one, that is:
(i,j∈[1,N-
m],j≠i);
(4) by ambiguity function _Define vectorWithSimilarityThat is:
Wherein ambiguity function _ (dm ij, n, r) and it is exponential function, n and r are respectively the gradient and width on exponential function boundary;
(5) defined function:
(6) similarly, step (2)~(5) are repeated, reconstruct generates one group of m+1 n dimensional vector n Q in order of sequencem+1(n,r);
(7) finally, fuzzy entropy can be calculated by following formula:
Fuzzy En (m, n, r, N)=ln [Qm(n,r)]-ln[Qm+1(n,r)]。
Part III, the tagsort based on LS-SVM:
There is hyperplane parameter selection and QuadraticProgramming (QP) in the application for SVM canonical algorithm
Matrix size is influenced very big, to cause solution scale excessive problem by training number of samples in problem solving.Suykens
J.A.K et al. proposes a kind of novel SVM, i.e. least square method supporting vector machine (Least Squares Support Vector
Machines,LS-SVM).LS-SVM compensates for the defects of SVM algorithm and deficiency, has obtained in two classification problems good
Using.
LS-SVM uses least square linear systematic error quadratic sum as loss function, in the target of its optimization problem
Two norms are used in function, and replace the inequality constraints condition in SVM canonical algorithm using equality constraint, so that LS-
The solution of the optimization problem of SVM method becomes the solution of the one group of system of linear equations obtained by Kuhn-Tucker condition, accelerates
Solving speed, and the computing resource needed for solving is less, achieves in pattern-recognition and the application of nonlinear function approximation
Good effect.Using LS-SVM as classifier in this patent, using radial basis function and Morlet wavelet function as LS-
The kernel function of SVM, to realize good J wave detection effect.
The mathematic(al) representation of the categorised decision function of LS-SVM are as follows:Wherein, K (z, zm) be LS-SVM kernel function, αmIndicate glug
Bright day multiplier, zmIndicate m-th of input vector, ωmIt is object vector, b is bias term.The expression formula of Radial basis kernel function is as follows
It is shown:Wherein, σ is nuclear parameter, can control the size of radial direction base core.Morlet kernel function
Expression formula it is as follows:Wherein, D is that input is special
The dimension of collection, parameter l indicate the scale factor of Morlet core.
Claims (1)
1. the J wave automatic testing method based on least square method supporting vector machine, it is characterised in that the following steps are included:
(1) ecg signal data collection data set is divided into training set, verifying collection and test set, training set, verifying collection and test set
In all include normal electrocardiosignal and the wave containing J electrocardiosignal, ecg signal data collection is pre-processed, i.e., to electrocardiosignal
Denoised, go baseline drift, remove artefact etc., and the heart is carried out to pretreated electrocardiosignal and claps segment processing;
(2) five layers of flexibly parsing wavelet decomposition then are carried out to the electrocardiosignal after segmentation, so that five straton band signals are obtained, spirit
Parsing small echo living has the wavelet basis of a pair of of Hilbert transform pairs, so as to neatly Control platform factor Q, redundancy R
And scale factor d;
(3) fuzzy entropy of this each layer of five stratons band signal is calculated, then using this five layers fuzzy entropy as electrocardiosignal
Feature vector, improvement of the fuzzy entropy as Sample Entropy algorithm, fuzzy entropy are blurred similarity measurement formula with an exponential function, make
Entropy must be obscured can be with Parameters variation transitions smooth;
(4) using the feature vector of ecg signal data concentration training collection as the input of least square method supporting vector machine classifier,
Least square method supporting vector machine is trained, and model selection, last test collection number are carried out by the feature vector of verifying collection
According to being tested, final mask is obtained;
(5) electrocardiosignal feature vector to be measured is input in trained least square method supporting vector machine final mask, thus
It realizes the Classification and Identification of normal electrocardiosignal Yu the electrocardiosignal of wave containing J, successfully detects J wave signal.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711037915.XA CN107837083B (en) | 2017-10-31 | 2017-10-31 | J wave automatic testing method based on least square method supporting vector machine |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711037915.XA CN107837083B (en) | 2017-10-31 | 2017-10-31 | J wave automatic testing method based on least square method supporting vector machine |
Publications (2)
Publication Number | Publication Date |
---|---|
CN107837083A CN107837083A (en) | 2018-03-27 |
CN107837083B true CN107837083B (en) | 2019-05-10 |
Family
ID=61681934
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201711037915.XA Active CN107837083B (en) | 2017-10-31 | 2017-10-31 | J wave automatic testing method based on least square method supporting vector machine |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107837083B (en) |
Families Citing this family (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109805928A (en) * | 2019-04-09 | 2019-05-28 | 太原理工大学 | A kind of cerebral apoplexy Method of EEG signals classification and system |
US20200390355A1 (en) * | 2019-06-11 | 2020-12-17 | Vios Medical, Inc. | System for detecting qrs complexes in an electrocardiography (ecg) signal |
CN111543981B (en) * | 2020-03-16 | 2023-04-18 | 浙江好络维医疗技术有限公司 | Dynamic electrocardiogram real-time filtering method based on segmented MODWT and adaptive threshold |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP2422696A1 (en) * | 2010-08-31 | 2012-02-29 | Marco Vettorello | Determining conditions of hypovolemia |
US8504142B2 (en) * | 2010-11-12 | 2013-08-06 | Oulun Yliopisto | Apparatus, method, and computer program for predicting risk for cardiac death |
AU2013204274A1 (en) * | 2012-04-23 | 2013-11-07 | Biosense Webster (Israel), Ltd. | Cardiac activation time detection |
AU2015200840A1 (en) * | 2014-02-26 | 2015-09-10 | Biosense Webster (Israel) Ltd. | Determination of reference annotation time from multi-channel electro-cardiogram signals |
JP6077343B2 (en) * | 2013-03-08 | 2017-02-08 | フクダ電子株式会社 | ECG data processing apparatus and control method thereof |
CN106503670A (en) * | 2016-11-03 | 2017-03-15 | 太原理工大学 | J ripples detection sorting technique based on correlation analysis feature selecting |
CN106650609A (en) * | 2016-10-26 | 2017-05-10 | 太原理工大学 | J-wave detection and classification method based on tunable Q-factor wavelet transform and higher-order cumulant |
-
2017
- 2017-10-31 CN CN201711037915.XA patent/CN107837083B/en active Active
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP2422696A1 (en) * | 2010-08-31 | 2012-02-29 | Marco Vettorello | Determining conditions of hypovolemia |
US8504142B2 (en) * | 2010-11-12 | 2013-08-06 | Oulun Yliopisto | Apparatus, method, and computer program for predicting risk for cardiac death |
AU2013204274A1 (en) * | 2012-04-23 | 2013-11-07 | Biosense Webster (Israel), Ltd. | Cardiac activation time detection |
JP6077343B2 (en) * | 2013-03-08 | 2017-02-08 | フクダ電子株式会社 | ECG data processing apparatus and control method thereof |
AU2015200840A1 (en) * | 2014-02-26 | 2015-09-10 | Biosense Webster (Israel) Ltd. | Determination of reference annotation time from multi-channel electro-cardiogram signals |
CN106650609A (en) * | 2016-10-26 | 2017-05-10 | 太原理工大学 | J-wave detection and classification method based on tunable Q-factor wavelet transform and higher-order cumulant |
CN106503670A (en) * | 2016-11-03 | 2017-03-15 | 太原理工大学 | J ripples detection sorting technique based on correlation analysis feature selecting |
Non-Patent Citations (3)
Title |
---|
Automated diagnosis of coronary artery disease using tunable-Q wavelet transform applied on heart rate signals;S. Patidar et al;《Knowledge-Based Systems》;20150731;摘要,正文附图,方法部分 * |
Decision Support System for Focal EEG Signals using;Rajeev Sharma, Mohit Kumar, Ram Bilas Pachori, U.;《Journal of Computational Science》;20170418;摘要,正文附图,方法部分 * |
互模糊熵的改进及其在心衰检测中的应用;吴学谦;《中国优秀硕士学位论文全文数据库医药卫生科技辑(月刊 )》;20141115;摘要,正文附图,方法部分 * |
Also Published As
Publication number | Publication date |
---|---|
CN107837083A (en) | 2018-03-27 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Smital et al. | Real-time quality assessment of long-term ECG signals recorded by wearables in free-living conditions | |
CN104473629B (en) | Automatic electrocardioelectrode placement error detection method based on kernel function classification algorithm | |
CN109117730B (en) | Real-time electrocardiogram atrial fibrillation judgment method, device and system and storage medium | |
Sun et al. | Morphological arrhythmia automated diagnosis method using gray-level co-occurrence matrix enhanced convolutional neural network | |
CN107837083B (en) | J wave automatic testing method based on least square method supporting vector machine | |
Xu et al. | Rule-based method for morphological classification of ST segment in ECG signals | |
CN109864714A (en) | A kind of ECG Signal Analysis method based on deep learning | |
CN112493995B (en) | Anesthesia state evaluation system and method suitable for patients of different ages | |
Malek et al. | Automated detection of premature ventricular contraction in ECG signals using enhanced template matching algorithm | |
CN109948396B (en) | Heart beat classification method, heart beat classification device and electronic equipment | |
CN108742697B (en) | Heart sound signal classification method and terminal equipment | |
CN103637795B (en) | Automatic diagnosis function detection method for electrocardiogram instrument | |
Desai et al. | Decision support system for arrhythmia beats using ECG signals with DCT, DWT and EMD methods: A comparative study | |
CN112971795B (en) | Electrocardiosignal quality evaluation method | |
CN112704503B (en) | Electrocardiosignal noise processing method | |
Sabor et al. | Detection of the interictal epileptic discharges based on wavelet bispectrum interaction and recurrent neural network | |
Wang et al. | A pooling convolution model for multi-classification of ECG and PCG signals | |
CN106419884B (en) | A kind of rate calculation method and system based on wavelet analysis | |
CN108338777A (en) | A kind of pulse signal determination method and device | |
Kamozawa et al. | A Detection Method of Atrial Fibrillation from 24‐hour Holter‐ECG Using CNN | |
Arsene | Design of deep convolutional neural network architectures for denoising electrocardiographic signals | |
Yun-Mei et al. | The abnormal detection of electroencephalogram with three-dimensional deep convolutional neural networks | |
Ayushi et al. | A survey of ECG classification for arrhythmia diagnoses using SVM | |
CN113171102B (en) | ECG data classification method based on continuous deep learning | |
Rashidinejad et al. | Patient-adaptable intracranial pressure morphology analysis using a probabilistic model-based approach |
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 |