CN102389302B - Analysis method of dynamic characteristics of electrocardiosignal - Google Patents

Abstract

The invention relates to an analysis method of dynamic characteristics of electrocardiosignal, which comprises the following steps of: firstly, utilizing a surrogate data algorithm to carry out the dynamic characteristic recognition of collected signals; secondly, converting the recognized electrocardiosignal into a weighted network, and further capturing the difference between the dynamic characteristics of different types of electrocardiosignal in a framework of a complex network by a point intensity distribution map; and finally, defining a statistic Rs, and successfully distinguishing normal electrocardiosignal from atrial fibrillation electrocardiosignal based on the statistic. In the technical scheme provided by the invention, the classification accuracy of the electrocardisignal is improved, and the information contained in the electrocardisignal is deeply discovered.

Description

A kind of electrocardiosignal dynamic characteristic analytical method

Technical field

The present invention relates to a kind of analytical method of electrocardiosignal, relate in particular to a kind of analytical method of electrocardiosignal dynamic characteristic.

Background technology

Heart disease has become one of main killer who threatens human health, therefore, develops that the automatic analytical tool of electrocardiosignal is significant fast and effectively.Just be based on this needs, nearly decades, people proposed the method for many analysis electrocardiosignaies: the sequence hypothesis check algorithm, the complexity algorithm, analysis of spectrum, method of wavelet analysis, these methods all are to extract feature from the time domain of electrocardiosignal or frequency domain, according to these features the electrocardiosignal that collects are carried out discriminator.Wherein, certain methods has been used to automatic defibrillator, and has obtained good effect.The non-linear dynamic feature of the cardiac system that but these methods all do not relate in the electrocardiosignal to be implied.In this simultaneously, some research worker begin to utilize Nonlinear Dynamics to analyze electrocardiosignal, and phase space reconfiguration has been widely applied in the ECG Signal Analysis as the bridge between the fluctuation of observation signal system time and the dynamical system space characteristics.But unlike signal is different to the requirement of phase space reconfiguration parameter (reconstruct dimension and time delay), the way that all adopts identical parameters to be reconstructed to all signals is difficult to the geometry of cardiac system is opened fully at present, thereby has limited the raising of electrocardiosignal analysis precision in phase space.

Summary of the invention

The technical problem that the present invention solves is: a kind of analytical method of electrocardiosignal dynamic characteristic is provided, overcome prior art analyze the nicety of grading of electrocardiosignal not high, fail deep layer ground and excavate the technical problem of the information that electrocardiosignal contains.

Technical scheme of the present invention is: a kind of analytical method of electrocardiosignal dynamic characteristic is provided, comprises the steps:

Gather electrocardiosignal: gather electrocardiosignal with electrocardiogram acquisition equipment, be designated as:

{ x _i, i=1,2,3 ..., n; Wherein: i represents sampling number;

Electrocardiosignal is carried out the dynamic characteristic analysis: the network that electrocardiosignal is mapped as weighting, each electrocardiosignal of gathering is a data segment, the network of described weighting is with a node in the corresponding described network of each data segment, between a pair of node the weights on limit be this to the distance between the corresponding data segment of node, be set to cut-point with R wave-wave peak position target electrocardiosignal section be cut into some data segment { s ₁, s ₂... s _m, a node in the corresponding described complex network of each data segment; Then, define the simple distance between a pair of data segment, and this distance is corresponded in the network weights on limit between corresponding node; Obtain the node intensity distribution of gained network;

Step 300: judge electrocardiosignal, that is: the node intensity distribution that obtains is carried out Gauss curve fitting, define statistic R on this basis _s

Further technical scheme of the present invention is: in gathering the electrocardiosignal step, also comprise and adopt the data alternate algorithm that the electrocardiosignal dynamic characteristic is identified.

Further technical scheme of the present invention is: adopt the data alternate algorithm to carry out the identification of electrocardiosignal dynamic characteristic and comprise the steps:

Given null hypothesis: suppose { x _iSequence is stochastic signal;

Generate alternate data: with signal { x _i, i=1,2,3 ..., n generates alternate data at random

J=1,2 ..., T;

Select for use correlation dimension as statistic: to select to embed dimension d from small to large, calculate the different { x under the dimensions that embed respectively _iCorrelation dimension D ₀, J=1,2 ..., the correlation dimension D of T ₁, D ₂..., D _T, D ₁, D ₂..., D _TAverage With standard deviation D _VarAnd maximum D _MaxMinima D _Min

Dynamic characteristic identification: if D ₀Not in the interval $[\min (D_{\min },(\stackrel{\‾}{D}-D_{\mathrm{var}})),\max (D_{\max },(\stackrel{\‾}{D}+D_{\mathrm{var}}))]$ In, then refuse null hypothesis; If D ₀In the interval $[\min (D_{\min },(\stackrel{\‾}{D}-D_{\mathrm{var}})),\max (D_{\max },(\stackrel{\‾}{D}+D_{\mathrm{var}}))]$ In, then assert the hypothesis establishment.

Further technical scheme of the present invention is: electrocardiosignal is being carried out in the dynamic characteristic analytical procedure s between two data segments _iWith s _jBetween the determining of distance, adopt following formula:

$d_{\mathrm{ij}}=\underset{l=\mathrm{0,1},...l_j-l_i}{\min }\frac{l}{l_i}\underset{k=1}{\overset{l_i}{\mathrm{\Σ}}}\left|\right|s_{i,k,}-s_{j,k+l}\left|\right|$

Wherein: l _iBe s _iLength, l _jBe s _jLength, establish l _i≤ l _j, s _{I, k}, s _{J, k+l}Be respectively s _i, s _jK on the data segment, k+1 point.

Electrocardiosignal is carried out come resulting some intensity distribution of match with the Gaussian function model in the dynamic characteristic analysis in step, define statistic R on this basis _s:

$R_s=\frac{\underset{\mathrm{st}=0}{\overset{\mathrm{\ω}}{\mathrm{\Σ}}}\mathrm{pro}\left(\mathrm{st}\right)}{\underset{\mathrm{st}=\mathrm{\ω}}{\overset{\max \left({\left\{s{t}_i\right\}}_{i=1}^m\right)}{\mathrm{\Σ}}}\mathrm{pro}\left(\text{st}\right)}$

Wherein: ω is corresponding some intensity level of Gauss curve fitting peak of curve, and pro (st) is that gained network mid point intensity size is the statistical probability of st,

The maximum of expression gained network node mid point intensity, m is the node number of gained network.

Technique effect of the present invention is: the present invention relates to a kind of analytical method of electrocardiosignal dynamic characteristic, at first utilize the alternate data algorithm to carry out dynamic characteristic identification to gathering the signal that comes; Then, the electrocardiosignal that identifies is converted into the complex network of weighting, and then under the framework of complex network, utilizes the some intensity distribution to catch the difference of dynamic characteristic between dissimilar electrocardiosignaies; At last, define statistic Rs, and successfully normal electrocardiosignal and atrial fibrillation electrocardiosignal are distinguished according to this statistic.Technical solution of the present invention has improved the nicety of grading of electrocardiosignal, and more deep layer ground excavates the information that electrocardiosignal contains.

Description of drawings

Fig. 1 is flow chart of the present invention.

Fig. 2 is electrocardiosignal dynamic characteristic identification process figure of the present invention.

The specific embodiment

Below in conjunction with specific embodiment, technical solution of the present invention is further specified.

As shown in Figure 1, the specific embodiment of the present invention is: a kind of analytical method of electrocardiosignal dynamic characteristic comprises the steps:

Step 100; Gather electrocardiosignal, that is: gather electrocardiosignal with electrocardiogram acquisition equipment, be designated as:

{ x _i, i=1,2,3 ..., n; Wherein: i represents sampling number.

Specific implementation process is as follows: gather electrocardiosignal with electrocardiogram acquisition equipment, in the embodiment of the invention, adopt the rhythm and pace of moving things of the electrocardiogram acquisition equipment records heart that singly leads.

Step 200: electrocardiosignal is carried out the dynamic characteristic analysis: the network that electrocardiosignal is mapped as weighting, each electrocardiosignal of gathering is a data segment, the network of described weighting is with a node in the corresponding described network of each data segment, between a pair of node the weights on limit be this to the distance between the corresponding data segment of node, be set to cut-point with R wave-wave peak position target electrocardiosignal section be cut into some data segment { s ₁, s ₂... s _m, a node in the corresponding described complex network of each data segment; Then, define the simple distance between a pair of data segment, and this distance is corresponded in the network weights on limit between corresponding node; Obtain the node intensity distribution of gained network;

Specific implementation process is as follows: be set to cut-point with R wave-wave peak position target electrocardiosignal section is cut into some data segments, might as well be designated as { s ₁, s ₂... s _m, each data segment corresponds to a node in the complex network; The weights on limit are that this is to the distance between the corresponding data segment of node between a pair of node.The weights on limit, that is: in the network between two nodes distance reflected the distance of corresponding data section corresponding track in phase space.And s between two data segments here _i, s _jThe determining of distance, adopt following formula:

$d_{\mathrm{ij}}=\underset{l=\mathrm{0,1},...l_j-l_i}{\min }\frac{l}{l_i}\underset{k=1}{\overset{l_i}{\mathrm{\Σ}}}\left|\right|s_{i,k,}-s_{j,k+l}\left|\right|$

Wherein: l _iBe s _iLength, l _jBe s _jLength, establish l _i≤ l _j, s _{I, k}, s _{J, k+l}Be respectively s _i, s _jK on the data segment, k+1 point.

Above conversion method is mapped as the nearer node of network middle distance with near the track same unstable cycle rail in the phase space, thereby the dynamic characteristic that contains in the electrocardiosignal has been embedded in the topological structure of complex network.

Utilize the some intensity distributions to investigate the different of atrial fibrillation signal and normal electrocardiosignal then.The point strength formula of i node is:

${\mathrm{st}}_i=\underset{j=1}{\overset{m}{\mathrm{\Σ}}}{d}_{\mathrm{ij}}$

M and d _IjGiven identical in represented meaning and front.The statistical probability that corresponding some intensity level is st is:

$\mathrm{pro}\left(\mathrm{st}\right)=\frac{\underset{j=1}{\overset{m}{\mathrm{\Σ}}}\mathrm{\δ}(\mathrm{st}-{\mathrm{st}}_j)}{m}$

Here, $\mathrm{\δ}\left(x\right)=\left\{\begin{array}{cc}1& x=0\\ 0& x\≠0\end{array}\right.$

At last, come resulting some intensity distribution of match with the Gaussian function model, namely use shape as G _j(x)=a _jExp (((x-μ _j)/σ _j) ²) Gaussian function stack obtain fitting function

Wherein: the number of function in the M representation model, parameter a _j, μ _j, σ _jEstimated to obtain by the non-linear least square of classics.Determine corresponding some intensity level ω of matched curve peak value, define statistic R on this basis _s:

$R_s=\frac{\underset{\mathrm{st}=0}{\overset{\mathrm{\ω}}{\mathrm{\Σ}}}\mathrm{pro}\left(\mathrm{st}\right)}{\underset{\mathrm{st}=\mathrm{\ω}}{\overset{\max \left({\left\{s{t}_i\right\}}_{i=1}^m\right)}{\mathrm{\Σ}}}\mathrm{pro}\left(\text{st}\right)}$

Wherein: ω is corresponding some intensity level of Gauss curve fitting peak of curve, and pro (st) is that gained network mid point intensity size is the statistical probability of st,

The maximum of expression gained network node mid point intensity, m is the node number of gained network.

$R_s=\frac{\underset{\mathrm{st}=0}{\overset{\mathrm{\ω}}{\mathrm{\Σ}}}\mathrm{pro}\left(\mathrm{st}\right)}{\underset{\mathrm{st}=\mathrm{\ω}}{\overset{\max \left({\left\{s{t}_i\right\}}_{i=1}^m\right)}{\mathrm{\Σ}}}\mathrm{pro}\left(\text{st}\right)}$

Wherein: ω is corresponding some intensity level of Gauss curve fitting peak of curve.

In the specific embodiment, training and test sample book all adopt the data among the MIT-BIH data base that Massachusetts Institute Technology provides.

Preferred implementation of the present invention is: in gathering the electrocardiosignal step, also comprise and adopt the data alternate algorithm that the electrocardiosignal dynamic characteristic is identified, as shown in Figure 2, adopt data alternate algorithm identification electrocardiosignal dynamic characteristic to comprise the steps:

Step 101: given null hypothesis: suppose { x _iSequence is stochastic signal.

Step 102: generate alternate data, that is: with signal { x _i, i=1,2,3 ..., n generates alternate data at random

J=1,2 ..., T.Detailed process is as follows: this signal is upset order at random generate T (generally selecting T 〉=30) group alternate data, the steps include: at first, generate one group and { x _iIsometric random number

Secondly, will

Ascending order is arranged, and the footnote sequence after the note ordering is (k ₁, k ₂..., k _n); Then, order again

At last, repeat above process T time, obtain T group alternate data:

J=1,2 ..., T.

Step 103: select for use correlation dimension as statistic, that is: select to embed dimension d from small to large, calculate under the different embedding dimensions { x respectively _iCorrelation dimension D ₀,

J=1,2 ..., the correlation dimension D of T ₁, D ₂..., D _T, D ₁, D ₂..., D _TAverage

With standard deviation D _VarAnd maximum D _MaxMinima D _Min

Step 104: dynamic characteristic identification, that is: if D ₀Not in the interval $[\min (D_{\min },(\stackrel{\‾}{D}-D_{\mathrm{var}})),\max (D_{\max },(\stackrel{\‾}{D}+D_{\mathrm{var}}))]$ In, then refuse null hypothesis, that is: this signal has more rich dynamic characteristic information; If D ₀In the interval $[\min (D_{\min },(\stackrel{\‾}{D}-D_{\mathrm{var}})),\max (D_{\max },(\stackrel{\‾}{D}+D_{\mathrm{var}}))]$ In, then assert the hypothesis establishment, that is: this signal may be subjected to noise jamming bigger, if it is carried out the dynamic characteristic analysis, can cause erroneous judgement largely.

Technique effect of the present invention is: the present invention relates to a kind of analytical method of electrocardiosignal dynamic characteristic, at first utilize the alternate data algorithm to carry out dynamic characteristic identification to gathering the signal that comes; Then, the electrocardiosignal that identifies is converted into the complex network of weighting, and then under the framework of complex network, utilizes the some intensity distribution to catch the difference of dynamic characteristic between dissimilar electrocardiosignaies; At last, define statistic Rs, and successfully normal electrocardiosignal and atrial fibrillation electrocardiosignal are distinguished according to this statistic.Technical solution of the present invention has improved the nicety of grading of electrocardiosignal, and more deep layer ground excavates the information that electrocardiosignal contains.

Above content be in conjunction with concrete preferred implementation to further describing that the present invention does, can not assert that concrete enforcement of the present invention is confined to these explanations.For the general technical staff of the technical field of the invention, without departing from the inventive concept of the premise, can also make some simple deduction or replace, all should be considered as belonging to protection scope of the present invention.

Claims (3)

1. the analytical method of an electrocardiosignal dynamic characteristic comprises the steps:

Gather electrocardiosignal: gather electrocardiosignal with electrocardiogram acquisition equipment, be designated as:

{ x _i, i=1,2,3 ..., n; Wherein: i represents sampling number;

Electrocardiosignal is carried out the dynamic characteristic analysis: electrocardiosignal is mapped as the network of weighting, to the electrocardiosignal of each collection, is set to cut-point with R wave-wave peak position target electrocardiosignal section is cut into some data segment { s ₁, s ₂... s _m, a node in the corresponding described complex network of each data segment; Then, define the simple distance between a pair of data segment, this simple distance is s between two data segments _iWith s _jBetween distance, adopt following formula:

$d_{\mathrm{ij}}=\underset{l=\mathrm{0,1},...l_j-l_i}{\min }\frac{1}{l_i}\underset{k=1}{\overset{l_i}{\mathrm{\Σ}}}\left|\right|s_{i,k,}-s_{j,k+l}\left|\right|$

Wherein: l _iBe s _iLength, l _jBe s _jLength, establish l _i≤ l _j, s _{I, k}, s _{J, k+l}Be respectively s _i, s _jK on the data segment, k+l point;

And this distance corresponded in the network weights on limit between corresponding node; Obtain the some intensity distribution of gained network; The point intensity distribution that obtains is carried out Gauss curve fitting, define statistic R on this basis _s,

$R_s=\frac{\underset{\mathrm{st}=0}{\overset{\mathrm{\ω}}{\mathrm{\Σ}}}\mathrm{pro}\left(\mathrm{st}\right)}{\underset{\mathrm{st}=\mathrm{\ω}}{\overset{\max \left({\left\{{\mathrm{st}}_i\right\}}_{i=1}^m\right)}{\mathrm{\Σ}}}\mathrm{pro}\left(\mathrm{st}\right)}$

Wherein: ω is corresponding some intensity level of Gauss curve fitting peak of curve, and pro (st) is that gained network mid point intensity size is the statistical probability of st,

The maximum of expression gained network node mid point intensity, m is the node number of gained network.

2. according to the analytical method of the described electrocardiosignal dynamic characteristic of claim 1, it is characterized in that: in gathering the electrocardiosignal step, also comprise and adopt the data alternate algorithm that the electrocardiosignal dynamic characteristic is identified.

3. according to the analytical method of the described electrocardiosignal dynamic characteristic of claim 2, it is characterized in that: adopt the data alternate algorithm to carry out the identification of electrocardiosignal dynamic characteristic and comprise the steps:

Given null hypothesis: suppose { x _iSequence is stochastic signal;

Generate alternate data: with signal { x _i, i=1,2,3 ..., n generates alternate data at random

Select for use correlation dimension as statistic: to select to embed dimension d from small to large, calculate the different { x under the dimensions that embed respectively _iCorrelation dimension D ₀,

Correlation dimension D ₁, D ₂..., D _T, D ₁, D ₂..., D _TAverage

With standard deviation D _VarAnd maximum D _MaxMinima D _Min

Dynamic characteristic identification: if D ₀Not in the interval $[\min (D_{\min },\left(\stackrel{\‾}{D}{-D}_{\mathrm{var}}\right)),\max (D_{\max },(\stackrel{\‾}{D}+D_{\mathrm{var}}))]$ In, then refuse null hypothesis; If D ₀In the interval $[\min (D_{\min },\left(\stackrel{\‾}{D}{-D}_{\mathrm{var}}\right)),\max (D_{\max },(\stackrel{\‾}{D}+D_{\mathrm{var}}))]$ In, then assert the hypothesis establishment.

Images