CN108324264A - A kind of detection method and system of atrial fibrillation - Google Patents

A kind of detection method and system of atrial fibrillation Download PDF

Info

Publication number
CN108324264A
CN108324264A CN201810063007.6A CN201810063007A CN108324264A CN 108324264 A CN108324264 A CN 108324264A CN 201810063007 A CN201810063007 A CN 201810063007A CN 108324264 A CN108324264 A CN 108324264A
Authority
CN
China
Prior art keywords
atrial fibrillation
neural network
boxcount
point
waves
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
Application number
CN201810063007.6A
Other languages
Chinese (zh)
Inventor
李健
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
JIANGSU KONSUNG BIOMEDICAL TECHNOLOGY Co Ltd
Original Assignee
JIANGSU KONSUNG BIOMEDICAL TECHNOLOGY Co Ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by JIANGSU KONSUNG BIOMEDICAL TECHNOLOGY Co Ltd filed Critical JIANGSU KONSUNG BIOMEDICAL TECHNOLOGY Co Ltd
Priority to CN201810063007.6A priority Critical patent/CN108324264A/en
Publication of CN108324264A publication Critical patent/CN108324264A/en
Pending legal-status Critical Current

Links

Classifications

    • AHUMAN NECESSITIES
    • A61MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61BDIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00Measuring for diagnostic purposes; Identification of persons
    • A61B5/24Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
    • A61B5/316Modalities, i.e. specific diagnostic methods
    • A61B5/318Heart-related electrical modalities, e.g. electrocardiography [ECG]
    • AHUMAN NECESSITIES
    • A61MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61BDIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00Measuring for diagnostic purposes; Identification of persons
    • A61B5/24Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
    • A61B5/316Modalities, i.e. specific diagnostic methods
    • A61B5/318Heart-related electrical modalities, e.g. electrocardiography [ECG]
    • A61B5/346Analysis of electrocardiograms
    • A61B5/349Detecting specific parameters of the electrocardiograph cycle
    • A61B5/366Detecting abnormal QRS complex, e.g. widening
    • AHUMAN NECESSITIES
    • A61MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61BDIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00Measuring for diagnostic purposes; Identification of persons
    • A61B5/72Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7235Details of waveform analysis
    • A61B5/7264Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems
    • A61B5/7267Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems involving training the classification device
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00Computing arrangements based on biological models
    • G06N3/02Neural networks
    • G06N3/08Learning methods
    • G06N3/084Backpropagation, e.g. using gradient descent

Landscapes

  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Biophysics (AREA)
  • Cardiology (AREA)
  • Artificial Intelligence (AREA)
  • Biomedical Technology (AREA)
  • Heart & Thoracic Surgery (AREA)
  • Public Health (AREA)
  • Animal Behavior & Ethology (AREA)
  • Surgery (AREA)
  • Medical Informatics (AREA)
  • Veterinary Medicine (AREA)
  • Pathology (AREA)
  • Evolutionary Computation (AREA)
  • Mathematical Physics (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Software Systems (AREA)
  • General Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Computing Systems (AREA)
  • Computational Linguistics (AREA)
  • Fuzzy Systems (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Physiology (AREA)
  • Psychiatry (AREA)
  • Signal Processing (AREA)
  • Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)

Abstract

The present invention relates to a kind of detection method of atrial fibrillation and systems, the condition that the segment signal whether there is P waves is added while the irregular characteristic of phase between extracting this section of ECG signal RR of embodiment, solve the problems, such as that sinus arrhythmia signal can have flase drop, and enter BP neural network method and simple logical condition is replaced to judge atrial fibrillation signal, improve the accuracy rate that atrial fibrillation is detected when issue mesh between inputting RR is less.

Description

A kind of detection method and system of atrial fibrillation
Technical field
The present invention relates to a kind of detection methods of atrial fibrillation, while the invention further relates to a kind of detecting systems of atrial fibrillation.
Background technology
Atrial fibrillation is one of most common arrhythmia cordis, and incidence 0.3%~0.4% of being grown up occurs with the increase at age Rate is multiplied, and more than 75 years old person's incidence is close to 10%.The electric shock of normal heart is dynamic since sinoatrial node.Due to sinoatrial node position Intersection in atrium dextrum and superior vena cava passes to a left side so the excitement of sinoatrial node is transmitted to atrium dextrum first by interatrial tract Atrium forms the P waves on electrocardiogram.Excitement is transmitted to atrioventricular node along preceding middle bachman bundle.Since atrioventricular node conduction of velocity is slow, Form the PR sections on electrocardiogram, the phase between also referred to as PR.Excitement is formed through Xinier reservoir, left and right beam branch synchronous activition left and right ventricles downwards QRS complex.Atrial fibrillation electrocardiosignal has the characteristics that:P waves disappear, instead continuous irregular baseline fluctuation (f waves), QRS- T wave morphologies be it is supraventricular, also can be Chong Die with f waves, or because of indoor aberrant conduction, Premature Ventricular Beats, pre-excitation syndrome etc. and It changes, when non-medication, between 120~200bpm of ventricular rate, the phase is absolutely irregular between RR.
In the prior art, mostly according to atrial fibrillation electrocardiosignal have RR between phase, that is, two adjacent QRS waves distance absolutely not Regular characteristic, extraction embody phase irregular characteristic between RR, the logical condition of judgement atrial fibrillation are determined according to the characteristic, and then really Whether electric signal of feeling relieved is atrial fibrillation signal.But this method is for sinus arrhythmia signal (since the phase equally has between its RR Phase absolute irregular characteristic between RR), can there are problems that flase drop.And simply logical condition is used to judge atrial fibrillation, between RR When phase number is few, can there is a problem of that precision is not high.
Invention content
It is an object of the invention to solve the above-mentioned problems, provide a kind of detection method of high-precision atrial fibrillation and be System.
The present invention provides a kind of detection method of high-precision atrial fibrillation, includes the following steps:
Step S1:Obtain ecg signal data, including the distance RR [i] of two adjacent QRS waves and corresponding each heart Whether the heart beat type fought is ventricular premature beat pvc [i], is then i=1, otherwise i=0 and corresponding each heartbeat whether there is P waves are denoted as P [i], and it is then 1 to exist, and there is no be then 0;
Step S2:It is ventricular premature beat, then more preset premature beat number threshold value NThr and room if the heartbeat type Property premature beat sum numpvc, if total premature ventricular contractions numpvc be less than premature beat number threshold value NThr, enter step S3;
Step S3:The first-order difference for calculating two adjacent QRS waves is less than the number of preset differential threshold thr numdRR;
Calculate the grid number boxcount occupied by reconstruction point;
Calculate the dispersion d that point (RR [i], RR [i+1]) is distributed relative to straight line y=x;
It calculates the distance dlineRR [i] in the point of (RR [i], RR [i+1]) to straight line y=x and is more than pre-determined distance threshold value The number numd of the point of ThrD;
There are the number nump of P waves for calculating;
Step S4:It will obtain described numdRR, oxcount, d, the numd, god of the nump as BP neural network input layer Through member, [1,0], the neuron of [0,1] as output layer, it is numhidden to take hidden layer neuron number, and input layer is to hiding The transmission function of layer is f (n), and the transmission function of hidden layer to output layer is g (n), and training obtains single hidden layer BP neural network Net { W, W ', B ', B " };
Step S5:The neural network net { W, W ', B ', B " } obtained in S4 is called to obtain value Y, specific invocation step is as follows: Enable x1=numdRR, x2=boxcount, x3=d, x4=numd, x5=nump inputs neural network net as input data, Obtain Y, wherein Y=[y [1], y [2]], y [i] calculation formula:
Step S6:Y value is obtained, if Y, which is [1,0], is determined as atrial fibrillation signal, if it is non-atrial fibrillation signal that Y, which is [0,1],.
Each heartbeat is denoted as P [i] with the presence or absence of P waves in step S1,(i=1 ..., n), it is early The value range of threshold value of fighting NThr is n/2~3n/5;If the numpvc in the step is more than NThr, then it is determined as non-atrial fibrillation.
Phase first-order difference is as follows less than the number numdRR concrete modes of thr between calculating RR in step S3:
Wherein
DRR [i]=RR [i+1]-RR [i]
Wherein, ranging from 0.06~0.12s of differential threshold thr.
The calculation of grid number boxcount is as follows in step S4:
It is used as x coordinate axis, RR [i+1] to be used as y-coordinate axis RR [i], draws reconstruction point, wherein i=1 ..., n-1;
It is covered with the grid of 20*20, wherein grid interval is 100ms;
Calculate the grid number boxcount occupied by reconstruction point.
Dispersion d calculation formula are as follows in step S4:
Wherein, std is the standard deviation of { dlineRR [i] }
DlineRR [i] is the distance that point (RR [i], RR [i+1]) arrives straight line y=x: 0<i<N-1,
M is the mean value of { dlineRR [i] }:
The present invention also provides a kind of detecting system of atrial fibrillation, including electrocardiosignal acquiring unit, ventricular premature beat are more single Member, computing unit, BP neural network training unit and neural network call unit, wherein electrocardiosignal acquiring unit obtain letter Breath includes:Whether the distance RR [i] of two adjacent QRS waves and the heart beat type of corresponding each heartbeat are ventricular premature beat pvc [i] is then i=1, and otherwise i=0 and corresponding each heartbeat are denoted as P [i] with the presence or absence of P waves, and it is 1 to exist then, is not present It is then 0;
Ventricular premature beat comparing unit, more preset premature beat threshold value NThr and total premature ventricular contractions numpvc, if room property is early The total numpvc that fights is less than threshold value NThr, then calls computing unit;
Computing unit, the first-order difference for calculating two adjacent QRS waves are less than thr number of preset differential threshold numdRR;
Calculate the grid number boxcount occupied by reconstruction point;
Calculate the dispersion d that point (RR [i], RR [i+1]) is distributed relative to straight line y=x;
The distance dlineRR [i] calculated in the point of (RR [i], RR [i+1]) to straight line y=x is big
In the number numd of the point of ThrD;
There are the number nump of P waves for calculating;
BP neural network training unit will obtain described numdRR, boxcount, d, numd, nump as BP nerve nets The neuron of network input layer, [1,0], the neuron of [0,1] as output layer, it is numhidden to take hidden layer neuron number, The transmission function of input layer to hidden layer is f (n), and the transmission function of hidden layer to output layer is g (n), and training obtains single hide Layer BP neural network
Net { W, W ', B ', B " };
Neural network call unit, call neural network net that the BP neural network training unit obtains W, W ', B ', B " } value Y is obtained, specific invocation step is as follows:Enable x1=numdRR, x2=boxcount, x3=d, x4=numd, x5= Nump inputs neural network net, obtains Y, wherein Y=[y [1], y [2]], y [i] calculation formula as input data:Y value is obtained, if Y, which is [1,0], is determined as atrial fibrillation signal, It is non-atrial fibrillation signal if Y is [0,1].
Each heartbeat is denoted as P [i] with the presence or absence of P waves in electrocardiosignal acquiring unit,(i= 1 ..., n), the value range of premature beat threshold value NThr is n/2~3n/5;If the numpvc in the ventricular premature beat comparing unit is big In NThr, then it is determined as non-atrial fibrillation.
Number numdRR of the phase first-order difference less than preset differential threshold thr is specifically square between calculating RR in computing unit Formula is as follows:
Wherein
DRR [i]=RR [i+1]-RR [i]
Wherein, ranging from 0.06~0.12s of differential threshold thr.
The calculation of grid number boxcount is as follows in computing unit:
It is used as x coordinate axis, RR [i+1] to be used as y-coordinate axis RR [i], draws reconstruction point, wherein i=1 ..., n-1;
It is covered with the grid of 20*20, wherein grid interval is 100ms;
Calculate the grid number boxcount occupied by reconstruction point.
Dispersion d calculation formula are as follows in computing unit:
Wherein, std is the standard deviation of { dlineRR [i] }
DlineRR [i] is the distance that point (RR [i], RR [i+1]) arrives straight line y=x: 0<i<N-1,
M is the mean value of { dlineRR [i] }:
Whether the present invention adds the segment signal while the irregular characteristic of phase between extracting this section of ECG signal RR of embodiment There are the condition of P waves, solve sinus arrhythmia signal (due between its RR the phase equally with the absolutely irregular spy of phase between RR Property) can there are problems that flase drop.And enter BP neural network method and simple logical condition is replaced to judge atrial fibrillation signal, it improves The accuracy rate of atrial fibrillation is detected when issue mesh between inputting RR is less.
To make the foregoing features and advantages of the present invention clearer and more comprehensible, special embodiment below, and coordinate institute's accompanying drawings It is described in detail below, wherein identical label indicates same or like unit or step.
Description of the drawings
The present invention is described in detail by following specific implementation mode and attached drawing for ease of explanation,.
Fig. 1 is a kind of flow diagram of the detection method of atrial fibrillation of the present invention;
Fig. 2 is reconstruction point distribution schematic diagram;
Fig. 3 is the relational graph of point (RR [i], RR [i+1]) and straight line y=x;
Fig. 4 is BP neural network structure chart;
Fig. 5 is a kind of schematic diagram of the detecting system of atrial fibrillation.
Specific implementation mode
In order to make the purpose , technical scheme and advantage of the present invention be clearer, with reference to the accompanying drawings and embodiments, right The present invention is further elaborated.
The present invention proposes that a kind of detection method of atrial fibrillation, overall flow figure please refer to Fig. 1,
Step 101:The distance of phase, that is, two adjacent QRS waves between ecg signal data, including RR is obtained, it is corresponding each Whether the heart beat type of the heartbeat i.e. heartbeat is that ventricular premature beat heartbeat and corresponding each heartbeat whether there is P.
The distance for obtaining phase, that is, two adjacent QRS waves between input data, including RR is denoted as RR [i], corresponding each Whether the heart beat type of the heartbeat i.e. heartbeat is ventricular premature beat heartbeat, is denoted as pvc [i], And corresponding each heartbeat is denoted as P [i] with the presence or absence of P waves, P [i]=0 without P waves, 1, there are P waves), wherein i=1 ..., n. In the case where CPU and memory allow, n can be larger with value, and the bigger algorithm accuracy of value of n is higher, but has in condition In the case of limit, the value of n is not easy too greatly.
Step 102:Calculate the total numpvc for including ventricular premature beat in input data.Formula is as follows:
More preset premature beat threshold value NThr and obtained numpvc, if numpvc>NThr is determined as non-atrial fibrillation, if Numpvc≤NThr, into next step.Wherein bigger number of the explanation comprising ventricular premature beat of NThr values is more, otherwise NThr is got over Small number of the explanation comprising ventricular premature beat is fewer, and the value range of General N Thr is n/2~3n/5.
Step 103:Phase first-order difference is less than the number numdRR of preset differential threshold thr, calculation formula between calculating RR It is as follows:
Wherein
DRR [i]=RR [i+1]-RR [i]
Wherein differential threshold thr embodies phase irregular degree between RR, and the phase, more irregularly thr took between the bigger RR of thr values The phase is more regular between being worth smaller RR, and usual differential threshold thr takes 0.06~0.12s.
Calculate the grid number boxcount occupied by reconstruction point.It is used as x coordinate axis, RR [i+1] to be used as y-coordinate RR [i] Axis draws reconstruction point, wherein i=1 ..., n-1;It is covered such as Fig. 2 with the grid (grid interval 100ms) of 20*20, in turn The grid number boxcount occupied by reconstruction point is calculated, such as boxcount=12 in Fig. 2;
The dispersion d that point (RR [i], RR [i+1]) is distributed relative to straight line y=x is calculated, calculation formula is as follows:
Wherein, std is the standard deviation of { dlineRR [i] }
DlineRR [i] is the distance that point (RR [i], RR [i+1]) arrives straight line y=x:
M is the mean value of { dlineRR [i] }:
The distance dlineRR [i] in the point of (RR [i], RR [i+1]) to straight line y=x is calculated more than distance threshold ThrD's The number numd of point.
Specific formula for calculation is as follows:
Wherein
Such as numd=11 in Fig. 3.Phase difference is bigger between two adjacent RR of the bigger explanations of wherein distance threshold ThrD, between RR Phase, more irregularly phase difference was smaller between two adjacent RR of the smaller explanations of distance threshold ThrD on the contrary, and the phase is more regular between RR.Distance Threshold value ThrD takes 40ms~120ms.
It calculates there are the number nump of P waves, formula is as follows:
The sequence for the step of wherein calculating can be interchanged, and changing its sequencing does not influence output result.
Step 104:Training BP neural network obtains net { W, W ', B ', B " }.The numdRR that step 103 is obtained, The neuron of boxcount, d, numd, nump as BP neural network input layer, [1,0], the nerve of [0,1] as output layer Member, it is numhidden to take hidden layer neuron number, and the transmission function of input layer to hidden layer is f (n), hidden layer to output The transmission function of layer is g (n), and training obtains single hidden layer BP neural network net { W, W ', B ', B " } (training of BP neural network Data come from MIT database afdb).The structure of BP neural network such as Fig. 4.Wherein numhidden takes 5~200, numhidden Accuracy with algorithm is in non-linear relation.F (n) is desirable Purelin (n)=n;G (n) is desirablePurelin (n)=n,
Step 105:The neural network net { W, W ', B ', B " } obtained in being walked in calling obtains value Y.Specific invocation step is such as Under:Enable x1=numdRR, x2=boxcount, x3=d, x4=numd, x5=nump inputs neural network as input data Net obtains Y, wherein Y=[y [1], y [2]], y [i] calculation formula:
Step 106:Atrial fibrillation signal is determine whether according to the Y value that step 105 obtains, if Y, which is [1,0], is determined as room Quiver signal, if it is non-atrial fibrillation signal that Y, which is [0,1],.
As shown in figure 5, the present invention also provides a kind of detecting system of atrial fibrillation, including electrocardiosignal acquiring unit, room property Premature beat comparing unit, computing unit, BP neural network training unit and neural network call unit, wherein electrocardiosignal obtain Unit obtains information:Whether the distance RR [i] of two adjacent QRS waves and the heart beat type of corresponding each heartbeat are room Property premature beat pvc [i] is then i=1, and otherwise i=0 and corresponding each heartbeat are denoted as P [i] with the presence or absence of P waves, exist, are 1, there is no be then 0;
Ventricular premature beat comparing unit, more preset premature beat threshold value NThr and total premature ventricular contractions numpvc, if room property is early The total numpvc that fights is less than premature beat threshold value NThr;
Computing unit, the first-order difference for calculating two adjacent QRS waves are less than the preset differential threshold thr Number numdRR;
Calculate the grid number boxcount occupied by reconstruction point;
Calculate the dispersion d that point (RR [i], RR [i+1]) is distributed relative to straight line y=x;
It calculates the distance dlineRR [i] in the point of (RR [i], RR [i+1]) to straight line y=x and is more than preset distance threshold The number numd of the point of ThrD;
There are the number nump of P waves for calculating;
BP neural network training unit will obtain described numdRR, boxcount, d, numd, nump as BP nerve nets The neuron of network input layer, [1,0], the neuron of [0,1] as output layer, it is numhidden to take hidden layer neuron number, The transmission function of input layer to hidden layer is f (n), and the transmission function of hidden layer to output layer is g (n), and training obtains single hide Layer BP neural network
Net { W, W ', B ', B " };
Neural network call unit, call neural network net that the BP neural network training unit obtains W, W ', B ', B " } value Y is obtained, specific invocation step is as follows:Enable x1=numdRR, X2=boxcount, x3=d, x4=numd, x5= Nump inputs neural network net, obtains Y, wherein Y=[y [1], y [2]], y [i] calculation formula as input data:Y value is obtained, if Y, which is [1,0], is determined as atrial fibrillation signal, It is non-atrial fibrillation signal if Y is [0,1].
Each heartbeat is denoted as P [i] with the presence or absence of P waves in electrocardiosignal acquiring unit,(i= 1 ..., n), the value range of premature beat threshold value NThr is n/2~3n/5;If the numpvc in the ventricular premature beat comparing unit is big In NThr, then it is determined as non-atrial fibrillation.
Phase first-order difference is as follows less than the number numdRR concrete modes of differential threshold thr between calculating RR in computing unit:
Wherein
DRR [i]=RR [i+1]-RR [i]
Wherein, ranging from 0.06~0.12s of differential threshold thr.
The calculation of grid number boxcount is as follows in computing unit:
It is used as x coordinate axis, RR [i+1] to be used as y-coordinate axis RR [i], draws reconstruction point, wherein i=1 ..., n-1;
It is covered with the grid of 20*20, wherein grid interval is 100ms;
Calculate the grid number boxcount occupied by reconstruction point.
Dispersion d calculation formula are as follows in computing unit:
Wherein, gtd is the standard deviation of { dlineRR [i] }
DlineRR [i] is the distance that point (RR [i], RR [i+1]) arrives straight line y=x: 0<i<N-1,
M is the mean value of { dlineRR [i] }:
The present invention is added with the presence or absence of P waves as one of neuron, effective to distinguish sinus arrhythmia and atrial fibrillation signal. BP neural network method is introduced by the irregular feature of phase between RR (numdRR, boxcount, d, numd) and atrial fibrillation signal without P The feature (nump) of wave is effectively combined, and the accuracy rate that atrial fibrillation detects when issue mesh is less between inputting RR is improved.
The above content is a further detailed description of the present invention in conjunction with specific preferred embodiments, and it cannot be said that The specific implementation of the present invention is confined to these explanations.For those of ordinary skill in the art to which the present invention belongs, exist Under the premise of not departing from present inventive concept, a number of simple deductions or replacements can also be made, all shall be regarded as belonging to the present invention's Protection domain.

Claims (10)

1. a kind of detection method of atrial fibrillation, which is characterized in that include the following steps:
Step S1:Obtain ecg signal data, including the distance RR [i] of two adjacent QRS waves and corresponding each heartbeat Whether heart beat type is ventricular premature beat pvc [i], is that then i=1, otherwise i=0 and corresponding each heartbeat whether there is P waves It is denoted as P [i], it is then 1 to exist, and there is no be then 0;
Step S2:It is that ventricular premature beat, then more preset premature beat number threshold value NThr and room property are early if the heartbeat type Fight total numpvc, if total premature ventricular contractions numpvc is less than premature beat number threshold value NThr, enters step S3;
Step S3:The first-order difference for calculating two adjacent QRS waves is less than the number of preset differential threshold thr numdRR
Calculate the grid number boxcount occupied by reconstruction point;
Calculate the dispersion d that point (RR [i], RR [i+1]) is distributed relative to straight line y=x;
It calculates the distance dlineRR [i] in the point of (RR [i], RR [i+1]) to straight line y=x and is more than preset distance threshold ThrD Point number numd;
There are the number nump of P waves for calculating;
Step S4:To obtain described boxcount, d, numd, neurons of the nump as BP neural network input layer, [1, 0], the neuron of [0,1] as output layer, it is numhidden, the transmission of input layer to hidden layer to take hidden layer neuron number Function is f (n), and the transmission function of hidden layer to output layer is g (n), it is trained obtain single hidden layer BP neural network net W, W ', B′,B″};
Step S5:The neural network net { W, W ', B ', B " } obtained in S4 is called to obtain value Y, specific invocation step is as follows:Enable x1 =numdRR, x2=boxcount, x3=d, x4=numd, x5=nump inputs neural network net, obtains as input data Y, wherein Y=[y [1], y [2]], y [i] calculation formula:
Step S6:Y value is obtained, if Y, which is [1,0], is determined as atrial fibrillation signal, if it is non-atrial fibrillation signal that Y, which is [0,1],.
2. a kind of detection method of atrial fibrillation as described in claim 1, which is characterized in that whether each heartbeat in the step S1 There are P waves to be denoted as P [i],The value range of the NThr be n/2~ 3n/5;If the numpvc in the step S2 is more than NThr, then it is determined as non-atrial fibrillation.
3. a kind of detection method of atrial fibrillation as described in claim 1, which is characterized in that calculate the phase one between RR in the step S3 Order difference is as follows less than the number numdRR concrete modes of thr:
Wherein
DRR [i]=RR [i+1]-RR [i]
Wherein, ranging from 0.06~0.12s of the differential threshold thr.
4. a kind of detection method of atrial fibrillation as described in claim 1, which is characterized in that grid number in the step S4 The calculation of boxcount is as follows:
It is used as x coordinate axis, RR [i+1] to be used as y-coordinate axis RR [i], draws reconstruction point, wherein i=1 ..., n-1;
It is covered with the grid of 20*20, wherein grid interval is 100ms;
Calculate the grid number boxcount occupied by reconstruction point.
5. a kind of detection method of atrial fibrillation as described in claim 1, which is characterized in that dispersion d is calculated in the step S4 Formula is as follows:
Wherein, std is the standard deviation of { dlineRR [i] }dlineRR[i] It is the distance that point (RR [i], RR [i+1]) arrives straight line y=x:
M is the mean value of { dlineRR [i] }:
6. a kind of detecting system of atrial fibrillation, which is characterized in that including electrocardiosignal acquiring unit, ventricular premature beat comparing unit, meter Unit, BP neural network training unit and neural network call unit are calculated, wherein
The electrocardiosignal acquiring unit obtains information:The distance RR [i] of two adjacent QRS waves and corresponding each Whether the heart beat type of heartbeat is ventricular premature beat pvc [i], is then i=1, otherwise whether i=0 and corresponding each heartbeat are deposited It is denoted as P [i] in P waves, it is then 1 to exist, and there is no be then 0;
The ventricular premature beat comparing unit, more preset premature beat threshold value NThr and total premature ventricular contractions numpvc, if room property Premature beat sum numpvc is less than premature beat threshold value NThr, then calls the computing unit;
The computing unit, the first-order difference for calculating two adjacent QRS waves are less than the default differential threshold The number numdRR of thr;
Calculate the grid number boxcount occupied by reconstruction point;
Calculate the dispersion d that point (RR [i], RR [i+1]) is distributed relative to straight line y=x;
The distance dlineRR [i] in the point of (RR [i], RR [i+1]) to straight line y=x is calculated more than pre-determined distance threshold value ThrD's The number numd of point;
There are the number nump of P waves for calculating
The BP neural network training unit will obtain described numdRR, boxcount, d, numd, nump as BP god Neuron through network input layer, [1,0], the neuron of [0,1] as output layer take the hidden layer neuron number to be The transmission function of numhidden, input layer to hidden layer are f (n), and the transmission function of hidden layer to output layer is g (n), training Obtain single hidden layer BP neural network net { W, W ', B ', B " };
The neural network call unit, call neural network net that the BP neural network training unit obtains W, W ', B ', B " } value Y is obtained, specific invocation step is as follows:Enable x1=numdRR, x2=boxcount, x3=d, x4=numd, x5= Nump inputs neural network net, obtains Y, wherein Y=[y [1], y [2]], y [i] calculation formula as input data:Y value is obtained, if Y, which is [1,0], is determined as atrial fibrillation signal, It is non-atrial fibrillation signal if Y is [0,1].
7. being characterized in that a kind of detecting system of atrial fibrillation as claimed in claim 6, which is characterized in that the electrocardiosignal obtains Each heartbeat in unit is taken to be denoted as P [i] with the presence or absence of P waves,The morning The value range of threshold value of fighting NThr is n/2~3n/5;If the numpvc in the ventricular premature beat comparing unit is more than NThr, then sentence It is set to non-atrial fibrillation.
8. being characterized in that a kind of detecting system of atrial fibrillation as claimed in claim 6, which is characterized in that in the computing unit Phase first-order difference is as follows less than the number numdRR concrete modes of differential threshold thr between calculating RR:
Wherein
DRR [i]=RR [i+1]-RR [i]
Wherein, ranging from 0.06~0.12s of the differential threshold thr.
9. being characterized in that a kind of detecting system of atrial fibrillation as claimed in claim 6, which is characterized in that in the computing unit The calculation of grid number boxcount is as follows:
It is used as x coordinate axis, RR [i+1] to be used as y-coordinate axis RR [i], draws reconstruction point, wherein i=1 ..., n-1;
It is covered with the grid of 20*20, wherein grid interval is 100ms;
Calculate the grid number boxcount occupied by reconstruction point.
10. being characterized in that a kind of detecting system of atrial fibrillation as claimed in claim 6, which is characterized in that the computing unit Middle dispersion d calculation formula are as follows:
Wherein, std is the standard deviation of { dlineRR [i] }dlineRR[i] It is the distance that point (RR [i], RR [i+1]) arrives straight line y=x:
M is the mean value of { dlineRR [i] }:
CN201810063007.6A 2018-01-23 2018-01-23 A kind of detection method and system of atrial fibrillation Pending CN108324264A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201810063007.6A CN108324264A (en) 2018-01-23 2018-01-23 A kind of detection method and system of atrial fibrillation

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201810063007.6A CN108324264A (en) 2018-01-23 2018-01-23 A kind of detection method and system of atrial fibrillation

Publications (1)

Publication Number Publication Date
CN108324264A true CN108324264A (en) 2018-07-27

Family

ID=62925631

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201810063007.6A Pending CN108324264A (en) 2018-01-23 2018-01-23 A kind of detection method and system of atrial fibrillation

Country Status (1)

Country Link
CN (1) CN108324264A (en)

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109044348A (en) * 2018-08-08 2018-12-21 广州视源电子科技股份有限公司 atrial fibrillation detection device and storage medium
CN109770858A (en) * 2019-03-28 2019-05-21 广州视源电子科技股份有限公司 A kind of method for detecting abnormality of electrocardiogram, device, equipment and storage medium
CN109770893A (en) * 2019-03-08 2019-05-21 东南大学 The method and device of atrial fibrillation position are quickly positioned in a kind of Holter analysis system
CN112971798A (en) * 2021-02-10 2021-06-18 武汉心络科技有限公司 Electrocardiosignal processing method, electrocardiosignal processing device, electrocardiosignal processing terminal and storage medium
CN113499079A (en) * 2021-06-18 2021-10-15 南京信息工程大学 Atrial fibrillation detection method in electrocardiogram

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101785670A (en) * 2009-01-22 2010-07-28 陈跃军 Intelligent blurry electrocardiogram on-line analyzer system
CN104523266A (en) * 2015-01-07 2015-04-22 河北大学 Automatic classification method for electrocardiogram signals
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
US20170112401A1 (en) * 2015-10-27 2017-04-27 CardioLogs Technologies Automatic method to delineate or categorize an electrocardiogram
CN107169424A (en) * 2017-04-24 2017-09-15 南京数维康信息科技有限公司 Atrial fibrillation detection algorithm based on artificial neural network

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101785670A (en) * 2009-01-22 2010-07-28 陈跃军 Intelligent blurry electrocardiogram on-line analyzer system
CN104523266A (en) * 2015-01-07 2015-04-22 河北大学 Automatic classification method for electrocardiogram signals
US20170112401A1 (en) * 2015-10-27 2017-04-27 CardioLogs Technologies Automatic method to delineate or categorize an electrocardiogram
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
CN107169424A (en) * 2017-04-24 2017-09-15 南京数维康信息科技有限公司 Atrial fibrillation detection algorithm based on artificial neural network

Cited By (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109044348A (en) * 2018-08-08 2018-12-21 广州视源电子科技股份有限公司 atrial fibrillation detection device and storage medium
CN109770893A (en) * 2019-03-08 2019-05-21 东南大学 The method and device of atrial fibrillation position are quickly positioned in a kind of Holter analysis system
CN109770858A (en) * 2019-03-28 2019-05-21 广州视源电子科技股份有限公司 A kind of method for detecting abnormality of electrocardiogram, device, equipment and storage medium
CN109770858B (en) * 2019-03-28 2021-07-13 广州视源电子科技股份有限公司 Method, device and equipment for detecting abnormality of electrocardiogram and storage medium
CN112971798A (en) * 2021-02-10 2021-06-18 武汉心络科技有限公司 Electrocardiosignal processing method, electrocardiosignal processing device, electrocardiosignal processing terminal and storage medium
CN112971798B (en) * 2021-02-10 2023-05-30 武汉心络科技有限公司 Electrocardiosignal processing method, electrocardiosignal processing device, electrocardiosignal processing terminal and storage medium
CN113499079A (en) * 2021-06-18 2021-10-15 南京信息工程大学 Atrial fibrillation detection method in electrocardiogram
CN113499079B (en) * 2021-06-18 2023-12-08 南京信息工程大学 Atrial fibrillation detection method in electrocardiogram

Similar Documents

Publication Publication Date Title
CN108324264A (en) A kind of detection method and system of atrial fibrillation
AU2015301633B2 (en) Atrial fibrillation detection using ventricular rate variability
CN106963367B (en) Overall system and method for detecting region of interest
Faes et al. A method for quantifying atrial fibrillation organization based on wave-morphology similarity
US7657307B2 (en) Method of and apparatus for classifying arrhythmias using scatter plot analysis
CN106066933B (en) The method for obtaining atrial fibrillation identification artificial neural network weight value matrix
EP3113674A1 (en) Multi-level heart failure event detection
WO2014205310A2 (en) Techniques for predicting cardiac arrhythmias based on signals from leads of electrocardiography
CA2605511A1 (en) Method and apparatus for detection of tachyarrhythmia using cycle lengths
CN109077714B (en) Signal identification method, device, equipment and storage medium
CN104736049A (en) A system and method for detecting a characteristic in an ECG waveform
JP2017518789A (en) Automatic differential diagnosis of worsening heart failure
EP1706031A2 (en) Cardiac monitoring
WO2006118854A2 (en) Method and apparatus for detection of tachyarrhythmia using cycle lengths
JP2013521867A (en) Time-frequency noise detection of intracardiac electrogram
CN109077720B (en) Signal processing method, device, equipment and storage medium
CN109620206B (en) Atrial fibrillation artificial intelligence identification method and device with ectopic heartbeat judgment function
US20230346305A1 (en) Parameter determination method, parameter determination apparatus, storage medium, and electronic device
US20090227884A1 (en) System for analysis of electrograms
Qin et al. Mechanisms for the maintenance of ventricular fibrillation: the nonuniform dispersion of refractoriness, restitution properties, or anatomic heterogeneities?
Rhaman et al. Successive RR Interval Analysis of PVC with Sinus Rhythm Using Fractal Dimension, Poincaré plot and Sample Entropy Method‖
CN108778112A (en) Detect boost pulse
Wilson et al. Stochastic pacing inhibits spatially discordant cardiac alternans
CN112244866B (en) Atrial arrhythmia detection in cardiac medical devices
Mattioni et al. Performance of an automatic external cardioverter-defibrillator algorithm in the discrimination of supraventricular from ventricular tachycardia

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
RJ01 Rejection of invention patent application after publication

Application publication date: 20180727