CN108511055A - Ventricular premature beat identifying system and method based on Multiple Classifier Fusion and diagnostic rule - Google Patents
Ventricular premature beat identifying system and method based on Multiple Classifier Fusion and diagnostic rule Download PDFInfo
- Publication number
- CN108511055A CN108511055A CN201710106859.4A CN201710106859A CN108511055A CN 108511055 A CN108511055 A CN 108511055A CN 201710106859 A CN201710106859 A CN 201710106859A CN 108511055 A CN108511055 A CN 108511055A
- Authority
- CN
- China
- Prior art keywords
- pvc
- fusion
- data
- results
- grader
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Molecular Biology (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)
- Medical Treatment And Welfare Office Work (AREA)
Abstract
The invention discloses a kind of ventricular premature beat identifying system and method based on Multiple Classifier Fusion and diagnostic rule, the system comprises：Taxon, include the LCNN sort modules and RNN sort modules to independent process ECG data, LCNN sort modules include the first grader of m different structure, at least exporting m the first classification results, RNN sort modules include the second grader of n different structure, at least exporting n the second classification results；Integrated unit obtains fusion results the first classification results and the second classification results are carried out fusion decision according to fusion decision rule；Judgement unit, non-PVC data and PVC data to judge to integrated unit according to PVC pathological characters differentiate, obtain PVC recognition results.The present invention merges the classification results of two kinds of graders of LCNN and RNN, and incorporates PVC pathological characters, the method being combined with medical diagnosis on disease rule using machine learning, improves the whole classification performance and accuracy rate of PVC identifications.
Description
Technical field
The present invention relates to a kind of ventricular premature beat identifying systems more particularly to a kind of based on Multiple Classifier Fusion and diagnostic rule
Ventricular premature beat identifying system and recognition methods belong to medical electronics technical field.
Background technology
The area of computer aided of ventricular premature beat (Premature Ventricular Contraction, hereinafter referred to as PVC) is examined
Disconnected to have very important clinical meaning, it can make doctor free from the ecg analysis of magnanimity, mitigate doctor's work
It measures, to improve the diagnosis efficiency of doctor.Currently, there are two main classes for the identifying system of area of computer aided ventricular premature beat：It is a kind of
It is to be identified using certain grader；Another kind of is the pathological characters showed according to PVC, then uses rule-based reasoning
Differentiated.Preceding a kind of Diagnostic Think process for usually having ignored doctor；And latter class is although it is contemplated that the Diagnostic Think of doctor
Process, it require that some characteristic points, such as R waves, QRS wave boundary point etc. of extraction PVC in advance, and how accurately to extract this
A little characteristic points are also that researcher needs the problem considered emphatically.
Invention content
It is based on Multiple Classifier Fusion and the identification of the ventricular premature beat of diagnostic rule the main purpose of the present invention is to provide a kind of
System and method, to overcome deficiency in the prior art.
For realization aforementioned invention purpose, the technical solution adopted by the present invention includes：
The ventricular premature beat identifying system based on Multiple Classifier Fusion and diagnostic rule that an embodiment of the present invention provides a kind of, packet
It includes：
Taxon, including LCNN sort modules and RNN sort modules, the LCNN sort modules, RNN sort modules are used
With independent process ECG data, the LCNN sort modules include the first grader of m different structure, the m the
For one grader at least to export m the first classification results, the RNN sort modules include the second classification of n different structure
Device, for the n the second graders at least to export n the second classification results, m, n are positive integer；
Integrated unit, to m the first classification results for exporting taxon according to fusion decision rule and n second
Classification results carry out fusion decision to obtain fusion results, and the fusion results include non-PVC data and PVC data；
And judgement unit, at least non-PVC data and PVC to judge to integrated unit according to PVC pathological characters
Data are differentiated, PVC recognition results are obtained.
Among some exemplary embodiments, first grader uses lead convolutional neural networks (lead
Convolutional neural network, hereinafter referred to as LCNN).
Among some exemplary embodiments, second grader uses recurrent neural network (recurrent
Neural network, hereinafter referred to as RNN).
Among some exemplary embodiments, the ventricular premature beat identifying system further includes：Pretreatment unit, to original
Beginning electrocardiogram (Electrocardiogram, hereinafter referred to as ECG) signal inputs taxon after being pre-processed.
Further, the pretreatment unit includes filter, at least dry to remove baseline drift noise and/or power frequency
Disturb noise.
Among some exemplary embodiments, the integrated unit includes：
First Fusion Module, to according to addition fusion decision rule to m of the m the first graders outputs the
One classification results carry out fusion decision；
Second Fusion Module, to according to addition fusion decision rule to n of the n the second graders outputs the
Two classification results carry out fusion decision；
Third Fusion Module, to export the first Fusion Module, the second Fusion Module according to mean value fusion decision rule
Fusion results carry out fusion decision to obtain final fusion results.
The embodiment of the present invention additionally provides a kind of ventricular premature beat recognition methods based on Multiple Classifier Fusion and diagnostic rule,
Including：
Use the first grader processing ECG data of m different structure in LCNN sort modules to export m first
Classification results；
Use the second grader processing ECG data of n different structure in RNN sort modules to export n second point
Class result；
The m of the output the first classification results and n the second classification results are merged according to fusion decision rule
For decision to obtain fusion results, the fusion results include non-PVC data and PVC data；
The non-PVC data and PVC data judged to integrated unit according to PVC pathological characters differentiate, obtain PVC knowledges
Other result.
Compared with prior art, advantages of the present invention includes：
Ventricular premature beat identifying system and recognition methods provided by the invention based on Multiple Classifier Fusion and diagnostic rule, fully
The respective advantage of LCNN and RNN is considered, respectively using them as the base grader of integrated study, then by two kinds of graders
Classification results are merged, to obtain a relatively good PVC classification results；Some pathology for also having incorporated PVC simultaneously are special
Sign, the method being combined with medical diagnosis on disease rule using machine learning are improved the whole classification performance of PVC identifications, effectively carried
The high accuracy rate of PVC identifications.
Description of the drawings
Fig. 1 is the ventricular premature beat recognition methods based on Multiple Classifier Fusion and diagnostic rule selected using the present invention in embodiment
The flow diagram differentiated for original electrocardiographicdigital figure signal；
Fig. 2 is typical PVC electrocardiograms schematic diagram；
Fig. 3 is the flow chart that the pathological characters in the preferred embodiment of the present invention using PVC differentiate EPVC data；
Fig. 4 is the flow that the pathological characters in the preferred embodiment of the present invention using PVC differentiate Enon-PVC data
Figure.
Specific implementation mode
In view of deficiency in the prior art, inventor is able to propose the present invention's through studying for a long period of time and largely putting into practice
Technical solution.The inventive principle of the present invention mainly identifies PVC first with Multiple Classifier Fusion method, obtains non-PVC classes and PVC
Then class recycles some pathological characters of PVC to make respectively to predicting the non-PVC classes come and PVC classes after Multiple Classifier Fusion
Judgement again improves the accuracy rate of PVC identifications with this.It as follows will be to the technical solution, its implementation process and principle etc.
It is further explained.
The one side of the embodiment of the present invention provides a kind of to be known based on Multiple Classifier Fusion and the ventricular premature beat of diagnostic rule
Other system comprising：
Taxon, including LCNN sort modules and RNN sort modules, the LCNN sort modules, RNN sort modules are used
With independent process ECG data, the LCNN sort modules include the first grader of m different structure, the m the
For one grader at least to export m the first classification results, the RNN sort modules include the second classification of n different structure
Device, for the n the second graders at least to export n the second classification results, m, n are positive integer；
Integrated unit, to m the first classification results for exporting taxon according to fusion decision rule and n second
Classification results carry out fusion decision to obtain fusion results, and the fusion results include non-PVC data and PVC data；
And judgement unit, at least non-PVC data and PVC to judge to integrated unit according to PVC pathological characters
Data are differentiated, PVC recognition results are obtained.
Among some exemplary embodiments, first grader uses lead convolutional neural networks.
Among some exemplary embodiments, second grader uses recurrent neural network.
Among some exemplary embodiments, the ventricular premature beat identifying system further includes：Pretreatment unit, to original
Beginning ECG signal inputs taxon after being pre-processed.
Further, the pretreatment unit includes filter, at least to denoising, specifically includes removal baseline drift
The noises such as shifting, Hz noise.
Among some exemplary embodiments, the integrated unit includes：
First Fusion Module, to according to addition fusion decision rule to m of the m the first graders outputs the
One classification results carry out fusion decision；
Second Fusion Module, to according to addition fusion decision rule to n of the n the second graders outputs the
Two classification results carry out fusion decision；
Third Fusion Module, to export the first Fusion Module, the second Fusion Module according to mean value fusion decision rule
Fusion results carry out fusion decision to obtain final fusion results.
Among some exemplary embodiments, first Fusion Module carries out the formula of addition fusion decision rule use
For：
Wherein PLCNN-jIndicate that the fusion results of i the first classification results, i are the integer more than or equal to 2, tmjIt indicates by m
The probability value for belonging to jth class that a first classification results obtain, j=0 or 1, wherein 0 indicates non-PVC data, 1 indicates PVC numbers
According to.
Preferably, second Fusion Module carry out the formula that addition fusion decision rule uses for：
Wherein PRNN-jIndicate that the fusion results of g the second classification results, g are the integer more than or equal to 2, ynjIt indicates by n-th
The probability value for belonging to jth class that a second classification results obtain, j=0 or 1, wherein 0 indicates non-PVC data, 1 indicates PVC numbers
According to.
Preferably, the third Fusion Module carry out the formula that mean value fusion decision rule uses for：
Pj=(1/2) * (PLCNN-j+PRNN-j)
Wherein PjFor the final fusion results of the first grader and the second grader, j=0 or 1, wherein the 0 non-PVC numbers of expression
According to 1 indicates PVC data, if PjMore than 0.5, then the original electrocardiographicdigital figure signal is PVC data, otherwise the original electrocardiographicdigital figure
Signal is non-PVC data.
The terminal and R wave characteristics of the present invention involved Characteristics of electrocardiogram such as QRS wave when using PVC diagnostic rules
Point has extracted, and is not limited to the extracting method of QRS wave terminal and R waves.
The embodiment of the present invention another aspect provides based on Multiple Classifier Fusion and diagnostic rule ventricular premature beat identification
Method comprising：
Use the first grader processing ECG data of m different structure in LCNN sort modules to export m first
Classification results；
Use the second grader processing ECG data of n different structure in RNN sort modules to export n second point
Class result；
The m of the output the first classification results and n the second classification results are merged according to fusion decision rule
For decision to obtain fusion results, the fusion results include non-PVC data and PVC data；
The non-PVC data and PVC data judged to integrated unit according to PVC pathological characters differentiate, obtain PVC knowledges
Other result.
Three steps of process point that ventricular premature beat recognition methods using the present invention carries out original electrocardiographicdigital figure signal P differentiations are complete
At the first step carries out noise suppression preprocessing first, and second step integrated classification device LCNN and RNN, it is special that PVC pathology is respectively adopted in third step
It levies and is judged again predicting the PVC data come and non-PVC data (being denoted as non-PVC) after Multiple Classifier Fusion, referring to Fig. 1 institutes
Show, concrete processing procedure is：
1. pretreatment：
First, by noises such as the filtered device removal baseline drift of ECG signal, Hz noises.
2. Multiple Classifier Fusion
The part connection of LCNN and weights shared mechanism significantly reduce the complexity of network, reduce training parameter
Number, with very strong robustness and fault-tolerant ability.And RNN is then a kind of deep learning mould with storing memory function
Type, it considers the incidence relation between sample, its main feature is that the output result of network depends not only upon current input, and
It is also associated with past input.There is LCNN and RNN the feature that need not extract hand-designed, their assorting process to be
Completely automatic, their input is original input data, then obtains final classification results by training, test.
As shown in Figure 1, after being pre-processed, ECG signal carries out PVC points via m1 LCNN grader and n2 RNN grader respectively
Class, the wherein structure of this m1 LCNN grader and n2 RNN grader are all different, by each grader, respectively
To m1 classification results of m1 grader and n2 classification results of n2 grader, in fact each grader output
It is original probability value, can judges whether the ECG signal is PVC according to this original probability value.But in order to promote entirety
Classification performance, the present invention uses addition fusion decision rule respectively to m1 classification results and the progress of n2 classification results first
Decision is merged, two fusion results is obtained, is then again melted above-mentioned two fusion results using mean value fusion decision rule
Decision is closed, to obtain final fusion results.In order to further enhance whole classification performance, the present invention also uses PVC pathology
Feature respectively judges the non-PVC data and PVC data that judge after Multiple Classifier Fusion again.
3.PVC diagnostic rules
The Some features of PVC electrocardiograms are：1. the roomy deformity of QRS wave, QRS wave form is different from other morphologically normal
QRS wave；2. the height that R wave heights are apparently higher than or are clapped less than the non-PVC hearts；3. there are QRS-T waves ahead of time, the phase between RR
Phase between (when front center claps the distance that R waves wholeheartedly the clap R waves before) RR that is averaged less than front.It is illustrated in figure 2 the heart of PVC records
Electrograph schematic diagram, N indicate that the non-PVC hearts are clapped, and V indicates that the PVC hearts are clapped.Therefore phase, QRS wave width, QRS wave between present invention selection RR
Amplitude, QRS wave similarity are as characteristic parameter.
After Multiple Classifier Fusion differentiates, the data for being predicted as non-PVC classes and PVC classes are required to diagnose rule by PVC
Then judge again, be Enon-PVC by the data markers of non-PVC classes are predicted as by Multiple Classifier Fusion, and is predicted as PVC
The data markers of class are EPVC.EPVC classes data and Enon-PVC classes data are passed through flow chart shown in Fig. 3 and Fig. 4 and are made further respectively
Secondary differentiation.Wherein, the acquisition of QRS wave similarity is then the correlation by calculating current QRS wave and morphologically normal QRS wave
Coefficient, using relative coefficient as the similarity measurement of QRS wave.Shape of the average QRS wave width by calculating before front center is clapped
Then the normal QRS wave width of state is averaged acquisition.If ECG records the condition for meeting flow chart 3, which just has can
Can be PVC.
It is worth noting that：Since the present invention is two classification problems, usable specific (Sp), sensitivity (Se),
Accuracy rate (Acc) and overall target youden index γ measure the quality of classifying quality, and in general youden index is bigger, classification
The whole classification performance of system is better.The confusion matrix of two classification is as shown in table 1 below：
1 confusion matrix of table
Shown in then each index is defined as follows：
Acc=(TP+TN)/(TP+TN+FP+FN) (1)
Se=TP/ (TP+FN) (2)
Sp=TN/ (TN+FP) (3)
γ=Se+Sp-1 (4)
Below by way of several embodiments technical solution that present invention be described in more detail.However, selected embodiment is only
For illustrating the present invention, and do not limit the scope of the invention.
Embodiment 1
Data source used in the present embodiment is in Chinese angiocardiopathy database (CCDD databases, http://
58.210.56.164/ccdd/)。
(1) in order to carry out noise suppression preprocessing, ECG records the bandpass filtering for first passing around 0.5~40Hz；
(2) 35840 (being recorded containing 3112 PVC) pretreated ECG records is used to be used as training sample；And it will be other
141046 records (being recorded containing 2148 PVC) are for testing.All training samples are respectively input in 4 LCNN and 6
Independent parallel training is carried out in RNN, wherein this 4 LCNN are the select results from trained more LCNN models
The larger LCNN models of otherness between preferable and each model.Similarly, this 6 RNN are also from trained more RNN moulds
The larger RNN models of otherness between select result is preferable in type and each model.After study, test sample is divided equally
Independent test is not carried out by this 4 LCNN and 6 RNN respectively obtain 4 LCNN classification results and 6 RNN classification results.It
Output valve be probability value, use tmjIndicate that m-th of LCNN classification results obtain belong to jth class probability value (j=0,1,
Wherein 0 indicates non-PVC classes, and 1 indicates PVC classes), ynjIndicate the probability value for belonging to jth class that n-th of RNN classification results obtains.
In fact, having obtained a judgement about disease by each grader, i.e., if tmjOr ynjMore than 0.5, then
The sample is judged to PVC classes, is otherwise non-PVC classes, but in order to promote whole classification results, the present embodiment is not first to every
Output valve obtained by one grader makes a decision, but according to these output valves be respectively adopted addition fusion decision rule by these
Classification results carry out fusion decision.
The classification results of 4 LCNN of addition fusion decision rule pair are used to merge first, formula is as follows：
Wherein PLCNN-jIndicate the fusion results of 4 LCNN classification results, tmjExpression is obtained by m-th of LCNN classification results
The probability value for belonging to jth class, j=0 or 1, wherein 0 indicates non-PVC data, 1 indicates PVC data；
Similarly, to the classification results of 6 RNN carry out the formula that uses of fusion for：
Wherein PRNN-jIndicate the fusion results of 6 RNN classification results, ynjWhat expression was obtained by n-th of RNN classification results
Belong to the probability value of jth class, j=0 or 1, wherein 0 indicates non-PVC data, 1 indicates PVC data；
Use mean value fusion decision rule by two fusion results P againLCNN-jAnd PRNN-jMerged, use formula for：
Pj=(1/2) * (PLCNN-j+PRNN-j)
Wherein PjFor the final fusion results of LCNN graders and RNN graders, i.e. sample belongs to the probability of jth class, j=
0 or 1, wherein 0 indicates non-PVC data, 1 indicates PVC data.If PjMore than 0.5, then after two kinds of Multiple Classifier Fusions, sample category
In 1 class, that is, PVC data, otherwise sample belongs to 0 class i.e. non-PVC data.The result obtained after two kinds of Multiple Classifier Fusion decisions is such as
Shown in the following table 2：
2 Multiple Classifier Fusion of table identifies PVC results
The upper lateral of upper table confusion matrix indicates that truthful data, the left side longitudinally indicate prediction data.As shown in Table 2, two
After kind Multiple Classifier Fusion decision, the sensitivity of acquisition is also relatively low, and overall target γ is nor very high.In order to further enhance whole
Body classification performance, therefore inventor is considered as some pathological characters of PVC and comes to being predicted after Multiple Classifier Fusion
Enon-PVC data and EPVC data judge again.
(3) from table 2 it can be seen that predicting the Enon-PVC data come after Multiple Classifier Fusion shares 135212+458 item notes
Record, EPVC data share 3686+1690 item records.Then use some pathological characters of PVC respectively to EPVC data and Enon-
PVC data judge that their decision flow chart is as shown in Figure 3 and Figure 4 again, differentiate parameter employed in flow chart 3-4
Value is all in accordance with obtained by experience.Finally, after in conjunction with LCNN, RNN and PVC diagnostic rule, obtained PVC recognition results such as the following table 3
It is shown.
The final recognition results of table 3PVC
In conclusion being obtained in more than 14 ten thousand test datas of CCDD databases using the PVC identifying systems of the present embodiment
PVC accuracy rate be 98.01%, specificity 98.04%, sensitivity 96.32%.Compared with table 2, although using
The accuracy rate that LCNN is obtained with the RNN methods being combined is higher, but sensitivity and overall target are also relatively low, by LCNN, RNN
The each index obtained after some basic pathology features of PVC are combined is promoted, and sensitivity and overall target
The amplitude of promotion is also bigger, as can be seen from Table 3, the whole of PVC identifications is improved using Multiple Classifier Fusion and medical diagnosis on disease rule
Body classification performance.
It should be appreciated that above-described is only some embodiments of the present invention, it is noted that for the common of this field
For technical staff, under the premise of not departing from the concept of the present invention, other modification and improvement can also be made, these are all
It belongs to the scope of protection of the present invention.
Claims (11)
1. a kind of ventricular premature beat identifying system based on Multiple Classifier Fusion and diagnostic rule, it is characterised in that including：
Taxon, including LCNN sort modules and RNN sort modules, the LCNN sort modules, RNN sort modules are to only
Vertical processing ECG data, the LCNN sort modules include the first grader of m different structure, and the m is first point a
For class device at least to export m the first classification results, the RNN sort modules include the second grader of n different structure, institute
For n the second graders stated at least to export n the second classification results, m, n are positive integer；
Integrated unit, to m the first classification results for exporting taxon according to fusion decision rule and n second classification
As a result fusion decision is carried out to obtain fusion results, and the fusion results include non-PVC data and PVC data；
And judgement unit, at least non-PVC data and PVC data to judge to integrated unit according to PVC pathological characters
Differentiated, obtains PVC recognition results.
2. ventricular premature beat identifying system according to claim 1, it is characterised in that：First grader is rolled up using lead
Product neural network.
3. ventricular premature beat identifying system according to claim 1, it is characterised in that：Second grader is using recurrence god
Through network.
4. ventricular premature beat identifying system according to claim 1, it is characterised in that further include：Pretreatment unit, to right
Original electrocardiographicdigital figure signal inputs taxon after being pre-processed.
5. ventricular premature beat identifying system according to claim 4, it is characterised in that：The pretreatment unit includes filtering
Device, at least removing baseline drift noise and/or Hz noise noise.
6. ventricular premature beat identifying system according to claim 1, which is characterized in that the integrated unit includes：
First Fusion Module, to m first point according to addition fusion decision rule to the m the first grader output
Class result carries out fusion decision；
Second Fusion Module, to n second point according to addition fusion decision rule to the n the second grader output
Class result carries out fusion decision；
Third Fusion Module, to melt what the first Fusion Module, the second Fusion Module exported according to mean value fusion decision rule
It closes result and carries out fusion decision to obtain final fusion results.
7. ventricular premature beat identifying system according to claim 6, which is characterized in that first Fusion Module carries out addition
The formula that fusion decision rule uses for：
Wherein PLCNN-jIndicate that the fusion results of i the first classification results, i are the integer more than or equal to 2, tmjIt indicates by m-th the
What one classification results obtained belongs to the probability value of jth class, j=0 or 1, wherein 0 indicates non-PVC data, 1 indicates PVC data.
8. ventricular premature beat identifying system according to claim 7, which is characterized in that second Fusion Module carries out addition
The formula that fusion decision rule uses for：
Wherein PRNN-jIndicate that the fusion results of g the second classification results, g are the integer more than or equal to 2, ynjIt indicates by n-th the
What two classification results obtained belongs to the probability value of jth class, j=0 or 1, wherein 0 indicates non-PVC data, 1 indicates PVC data.
9. ventricular premature beat identifying system according to claim 8, which is characterized in that the third Fusion Module carries out mean value
The formula that fusion decision rule uses for：
Pj=(1/2) * (PLCNN-j+PRNN-j)
Wherein PjFor the final fusion results of the first grader and the second grader, j=0 or 1, wherein 0 indicates non-PVC data, 1
PVC data are indicated, if PjMore than 0.5, then the original electrocardiographicdigital figure signal is PVC data, otherwise the original electrocardiographicdigital figure signal
For non-PVC data.
10. ventricular premature beat identifying system according to claim 1, it is characterised in that：Differentiated in the judgement unit
Based on PVC pathological characters extracted, it is preferred that the PVC pathological characters include that phase, average QRS wave are wide between RR
Any one in degree, QRS wave starting and terminating point, QRS wave amplitude, QRS wave similarity or two or more combinations；It is preferred that
, the QRS wave similarity is obtained by calculating the relative coefficient of current QRS wave and morphologically normal QRS wave, and will be related
Similarity measurement of the property coefficient as QRS wave；The average QRS wave width is morphologically normal before front center is clapped by calculating
Then QRS wave width is averaged acquisition.
11. a kind of ventricular premature beat recognition methods based on Multiple Classifier Fusion and diagnostic rule, it is characterised in that including：
Use the first grader processing ECG data of m different structure in LCNN sort modules to export m first classification
As a result；
The second grader processing ECG data of n different structure in RNN sort modules is used to be tied to export n second classification
Fruit；
The m of the output the first classification results and n the second classification results are subjected to fusion decision according to fusion decision rule
To obtain fusion results, the fusion results include non-PVC data and PVC data；
The non-PVC data and PVC data judged to integrated unit according to PVC pathological characters differentiate, obtain PVC identification knots
Fruit.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710106859.4A CN108511055B (en) | 2017-02-27 | 2017-02-27 | Ventricular premature beat recognition system and method based on classifier fusion and diagnosis rules |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710106859.4A CN108511055B (en) | 2017-02-27 | 2017-02-27 | Ventricular premature beat recognition system and method based on classifier fusion and diagnosis rules |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108511055A true CN108511055A (en) | 2018-09-07 |
| CN108511055B CN108511055B (en) | 2021-10-12 |
Family
ID=63373034
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710106859.4A Active CN108511055B (en) | 2017-02-27 | 2017-02-27 | Ventricular premature beat recognition system and method based on classifier fusion and diagnosis rules |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108511055B (en) |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109350032A (en) * | 2018-10-16 | 2019-02-19 | 武汉中旗生物医疗电子有限公司 | A kind of classification method, system, electronic equipment and storage medium |
| CN109480824A (en) * | 2018-12-11 | 2019-03-19 | 武汉中旗生物医疗电子有限公司 | Processing method, device and the server of EGC waveform data |
| CN109907753A (en) * | 2019-04-23 | 2019-06-21 | 杭州电子科技大学 | A kind of various dimensions ECG signal intelligent diagnosis system |
| CN109949929A (en) * | 2019-03-19 | 2019-06-28 | 挂号网（杭州）科技有限公司 | A kind of assistant diagnosis system based on the extensive case history of deep learning |
| CN110037687A (en) * | 2019-04-09 | 2019-07-23 | 上海数创医疗科技有限公司 | Based on the ventricular premature beat heartbeat localization method and device for improving convolutional neural networks |
| CN110310745A (en) * | 2019-05-21 | 2019-10-08 | 上海交通大学医学院附属瑞金医院 | The therapeutic scheme recommender system that medical guide and data-driven combine |
| WO2020061987A1 (en) * | 2018-09-28 | 2020-04-02 | 深圳大学 | Multi-lead combined electrocardiogram classification method |
| CN111329469A (en) * | 2020-03-05 | 2020-06-26 | 广州天嵌计算机科技有限公司 | Arrhythmia prediction method |
| CN112971796A (en) * | 2021-02-09 | 2021-06-18 | 中山大学附属第三医院（中山大学肝脏病医院） | Myocardial infarction prediction device, system and storage medium |
| CN113116300A (en) * | 2021-03-12 | 2021-07-16 | 复旦大学 | Physiological signal classification method based on model fusion |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103284702A (en) * | 2013-06-17 | 2013-09-11 | 中国科学院苏州纳米技术与纳米仿生研究所 | Electrocardiogram and pulse wave relation analysis method and method and device of fusion analysis |
| CN104367317A (en) * | 2014-10-15 | 2015-02-25 | 北京理工大学 | Electrocardiogram electrocardiosignal classification method with multi-scale characteristics combined |
| WO2016038585A1 (en) * | 2014-09-12 | 2016-03-17 | Blacktree Fitness Technologies Inc. | Portable devices and methods for measuring nutritional intake |
| CN105530858A (en) * | 2013-09-16 | 2016-04-27 | 皇家飞利浦有限公司 | System and method for estimating cardiovascular fitness of a person |
| CN106214145A (en) * | 2016-07-20 | 2016-12-14 | 杨平 | A kind of electrocardiogram classification method based on degree of depth learning algorithm |
| US20170032221A1 (en) * | 2015-07-29 | 2017-02-02 | Htc Corporation | Method, electronic apparatus, and computer readable medium of constructing classifier for disease detection |
-
2017
- 2017-02-27 CN CN201710106859.4A patent/CN108511055B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103284702A (en) * | 2013-06-17 | 2013-09-11 | 中国科学院苏州纳米技术与纳米仿生研究所 | Electrocardiogram and pulse wave relation analysis method and method and device of fusion analysis |
| CN105530858A (en) * | 2013-09-16 | 2016-04-27 | 皇家飞利浦有限公司 | System and method for estimating cardiovascular fitness of a person |
| WO2016038585A1 (en) * | 2014-09-12 | 2016-03-17 | Blacktree Fitness Technologies Inc. | Portable devices and methods for measuring nutritional intake |
| CN104367317A (en) * | 2014-10-15 | 2015-02-25 | 北京理工大学 | Electrocardiogram electrocardiosignal classification method with multi-scale characteristics combined |
| US20170032221A1 (en) * | 2015-07-29 | 2017-02-02 | Htc Corporation | Method, electronic apparatus, and computer readable medium of constructing classifier for disease detection |
| CN106214145A (en) * | 2016-07-20 | 2016-12-14 | 杨平 | A kind of electrocardiogram classification method based on degree of depth learning algorithm |
Non-Patent Citations (3)
| Title |
|---|
| 周飞燕等: "卷积神经网络研究综述", 《计算机学报》 * |
| 樊承柱: "基于深度神经网络的心电图自动分类方法研究与实现", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
| 金林鹏等: "基于集成学习的临床心电图分类算法研究", 《生物医学工程学杂志》 * |
Cited By (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2020061987A1 (en) * | 2018-09-28 | 2020-04-02 | 深圳大学 | Multi-lead combined electrocardiogram classification method |
| CN109350032B (en) * | 2018-10-16 | 2022-03-18 | 武汉中旗生物医疗电子有限公司 | Classification method, classification system, electronic equipment and storage medium |
| CN109350032A (en) * | 2018-10-16 | 2019-02-19 | 武汉中旗生物医疗电子有限公司 | A kind of classification method, system, electronic equipment and storage medium |
| CN109480824A (en) * | 2018-12-11 | 2019-03-19 | 武汉中旗生物医疗电子有限公司 | Processing method, device and the server of EGC waveform data |
| CN109480824B (en) * | 2018-12-11 | 2021-10-08 | 武汉中旗生物医疗电子有限公司 | Method and device for processing electrocardio waveform data and server |
| CN109949929A (en) * | 2019-03-19 | 2019-06-28 | 挂号网（杭州）科技有限公司 | A kind of assistant diagnosis system based on the extensive case history of deep learning |
| CN110037687A (en) * | 2019-04-09 | 2019-07-23 | 上海数创医疗科技有限公司 | Based on the ventricular premature beat heartbeat localization method and device for improving convolutional neural networks |
| CN109907753A (en) * | 2019-04-23 | 2019-06-21 | 杭州电子科技大学 | A kind of various dimensions ECG signal intelligent diagnosis system |
| CN109907753B (en) * | 2019-04-23 | 2022-07-26 | 杭州电子科技大学 | Multi-dimensional ECG signal intelligent diagnosis system |
| CN110310745A (en) * | 2019-05-21 | 2019-10-08 | 上海交通大学医学院附属瑞金医院 | The therapeutic scheme recommender system that medical guide and data-driven combine |
| CN110310745B (en) * | 2019-05-21 | 2021-12-03 | 上海交通大学医学院附属瑞金医院 | Treatment plan recommendation system combining medical guide and data drive |
| CN111329469A (en) * | 2020-03-05 | 2020-06-26 | 广州天嵌计算机科技有限公司 | Arrhythmia prediction method |
| CN112971796A (en) * | 2021-02-09 | 2021-06-18 | 中山大学附属第三医院（中山大学肝脏病医院） | Myocardial infarction prediction device, system and storage medium |
| CN113116300A (en) * | 2021-03-12 | 2021-07-16 | 复旦大学 | Physiological signal classification method based on model fusion |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108511055B (en) | 2021-10-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108511055A (en) | Ventricular premature beat identifying system and method based on Multiple Classifier Fusion and diagnostic rule | |
| CN109543526B (en) | True and false facial paralysis recognition system based on depth difference characteristics | |
| Rubin et al. | Recognizing abnormal heart sounds using deep learning | |
| Kropf et al. | ECG classification based on time and frequency domain features using random forests | |
| CN109858540B (en) | Medical image recognition system and method based on multi-mode fusion | |
| CN109727243A (en) | Breast ultrasound image recognition analysis method and system | |
| CN108932452A (en) | Arrhythmia cordis beat classification method based on multiple dimensioned convolutional neural networks | |
| CN105997055A (en) | Automatic classification method, system and device of electrocardiosignal ST band | |
| CN108464840A (en) | A kind of breast lump automatic testing method and system | |
| CN110310282B (en) | System for intelligently analyzing corneal nerve fibers by using in-vivo confocal microscope images | |
| CN106980815A (en) | Facial paralysis objective evaluation method under being supervised based on H B rank scores | |
| CN110991536A (en) | Training method of early warning model of primary liver cancer | |
| CN107301409A (en) | Learn the system and method for processing electrocardiogram based on Wrapper feature selectings Bagging | |
| CN111462082A (en) | Focus picture recognition device, method and equipment and readable storage medium | |
| CN117116477A (en) | Construction method and system of prostate cancer disease risk prediction model based on random forest and XGBoost | |
| CN114469120A (en) | Multi-scale Dtw-BiLstm-Gan electrocardiosignal generation method based on similarity threshold migration | |
| CN112741651A (en) | Method and system for processing ultrasonic image of endoscope | |
| CN116864140A (en) | Intracardiac branch of academic or vocational study postoperative care monitoring data processing method and system thereof | |
| Wei et al. | Multi-feature fusion for ultrasound breast image classification of benign and malignant | |
| Taher et al. | Extraction of sputum cells using thresholding techniques for lung cancer detection | |
| CN114140830A (en) | Repeated identification inhibition method based on circulating tumor cell image | |
| CN112927808A (en) | Thyroid ultrasound image-based nodule grading system and method | |
| Arbab et al. | Automatic Detection and Classification of Acute Lymphoblastic Leukemia Using Convolution Neural Network | |
| Mangla et al. | Brain tumor detection and classification by MRI images using deep learning techniques | |
| Elzennary et al. | Early Deep Detection for Diabetic Retinopathy |
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 |