CN109887594A - A kind of multi-lead arrhythmia cordis intelligent diagnosing method based on MODWT and TDCNN - Google Patents
A kind of multi-lead arrhythmia cordis intelligent diagnosing method based on MODWT and TDCNN Download PDFInfo
- Publication number
- CN109887594A CN109887594A CN201811544778.3A CN201811544778A CN109887594A CN 109887594 A CN109887594 A CN 109887594A CN 201811544778 A CN201811544778 A CN 201811544778A CN 109887594 A CN109887594 A CN 109887594A
- Authority
- CN
- China
- Prior art keywords
- data
- lead
- tdcnn
- arrhythmia cordis
- modwt
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Landscapes
- Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)
Abstract
A kind of multi-lead arrhythmia cordis intelligent diagnosing method based on MODWT and TDCNN, comprising: the true ECG data R wave position of random length a) is found based on MODWT, data are split；B) it using MIT-BIH arrhythmia cordis database as training test data source, establishes based on TDCNN and trains neural network；C) intelligent diagnostics are carried out to true ECG data using trained network, and finds arrhythmia cordis position；The present invention can be split the true ECG data of random length according to different people R wave position, heart rate, and fully consider the inherent correlation between lead, it can be readily appreciated that easy to accomplish, precision of prediction is higher, is more suitable for the diagnosis of arrhythmia cordis.
Description
Technical field
The present invention relates to one kind based on MODWT (Maximal overlap discrete wavelet transform) and
The multi-lead arrhythmia cordis intelligent diagnostics of TDCNN (Two-dimensional convolutional neural networks)
Method belongs to electrocardiogram intelligent diagnostics field.
Background technique
ECG examination is the Common item of physical examination, if patient is doubtful arrhythmia cordis symptom, can generally also go to hospital
Do an electrocardiogram, still, a few minutes or more than ten minutes electrocardiogram in hospital, it is difficult to find the problem, at this time doctor can
It can give patient one ecg detector that can be dressed, usually to wear two weeks or for more time, this can generate several hundred small
When electrocardiogram, doctor wants inspection in one second one second, this is time consuming, if machine can first be allowed to carry out intelligent decision, looks for
The time location of possible arrhythmia cordis out, then further diagnosed by doctor for these places, it is clear that it can be greatly
It improves efficiency, economizes on resources, the invention original intention of this patent derives from this.
Currently, existing research person is studied for arrhythmia cordis intelligent diagnostics.Such as early stage utilization support vector machines,
The machine learning methods such as KNN diagnose arrhythmia cordis；In recent years with the rise of deep learning, some scholars propose utilization
The new method that deep learning diagnoses arrhythmia cordis: such as foreign countries Wu Enda leads machine learning group in Stamford to propose to use
CNN carries out the cardiac arrhythmia detection of cardiologist's grade, can identify 14 class cardiac arrhythmia types.But what this method was extracted is single
Leads, it is difficult to the types of arrhythmia that reply needs that multi-lead data is combined just to can be carried out diagnosis, meanwhile, this method is based on
Time (note: 1 second) is split diagnostic data, does not account for the different difference of heart rate between individual, when bradycardia,
Many segmentation data are difficult to include a complete heart beat cycle；Domestic high rock etc. is using CNN to the arrhythmia cordis intelligence of multi-lead
Diagnostic method is studied, this method difference different also without heart rate between consideration individual, meanwhile, with much channel communication
Handle multi-lead data, be virtually detached from the inner link between lead, in fact, the data source of multi-lead in electrocardio to
The reprojection of amount has contacting spatially between lead and lead.
Summary of the invention
It is an object of the invention to overcome the shortcomings of the prior art, and provide it is a kind of can according to different people R wave position,
Heart rate is split the true ECG data of random length, and fully considers the inherent correlation between lead, is easy to manage
Solution, easy to accomplish, precision of prediction is higher, is more suitable for the multi-lead arrhythmia cordis based on MODWT and TDCNN of arrhythmia cordis diagnosis
Intelligent diagnosing method.
To achieve the goals above, the technical solution adopted in the present invention is as follows: a kind of based on the more of MODWT and TDCNN
Lead arrhythmia cordis intelligent diagnosing method, it includes the following steps:
A) the true ECG data R wave position of random length is found based on MODWT, data is split；
B) it using MIT-BIH arrhythmia cordis database as training test data source, establishes based on TDCNN and trains nerve
Network；
C) intelligent diagnostics are carried out to true ECG data using trained network, and finds arrhythmia cordis position；Its
It is characterized in that:
The step a) finds the true ECG data R wave position of random length based on MODWT, is split to data
Method are as follows: be equipped with multi-lead ECG data X:
Wherein: m indicates lead number, and n indicates sampling number (because of sample frequency fcIt is known that can obtain at sampled point i to it is corresponding when
Carve ti=i/fc), include the following steps:
Step 1: taking the 1st leads as analysis data, obtain sampled point locating for continuous s R wave using MODWT programming
Position w1、w2、…、ws, it is seen that: 1 < w1<w2…<ws<n；
Step 2: obtain the interval RR:
y1=w2-w1,y2=w3-w2,…,ys-1=ws-ws-1
Taking is worth heart rate yz, for comprising a heart beat cycle, in w1Place: data intercept are as follows:
Note that when subscript<0 or>n when, give up the segment data.Data are normalized, and by stretching or pressing
Contracting transforms to fixed length N (note: can be set as the sample frequency of training data):
x111..., x11N
Wherein, subscript the 1st " 1 " indicates lead 1, the 1st heart beat cycle of the 2nd " 1 " expression, the 1st heart of the 3rd expression
The sampling point position of hop cycle；Similarly, other positions data intercept is obtained:
x121..., x12N, x131..., x13N..., x1s1..., x1sN。
Step 3: same to handle on the basis of the 1st lead R wave position to other leads；By each lead in each heart beat cycle
Fixed-length data be sequentially arranged above and below, obtain 2-D data divide and combine Y:
Y=[Y1..., Ys],
Wherein:
This is the heart beat cycle segmentation data that need to be diagnosed.
As preferred: the step b) is based on using MIT-BIH arrhythmia cordis database as training test data source
TDCNN is established and is trained neural network method are as follows:
Step 1): obtaining MIT-BIH arrhythmia cordis database, gives up exception lead and the few arrhythmia cordis class of data volume
Type selectes II lead, the normal sinus rhythm (N) of VI lead, left bundle branch block (L), right bundle branch block (R), room
Property premature beat (A) and five kinds of types of arrhythmia of ventricular premature beat (V) as analysis data；According to the R wave position of mark, heartbeat is obtained
Period divisions data；
Step 2): TDCNN is constructed according to such as the following table 1；
1 TDCNN structure of table
| Input layer | II lead, VI lead: 2 × 360 |
| Convolutional layer 1 | Filter quantity: 32, size: 2 × 9 |
| Pond layer 1 | Size: 2 × 2 |
| Convolutional layer 2 | Filter quantity: 32, size: 1 × 9 |
| Convolutional layer 2 | Size: 1 × 2 |
| Full articulamentum | Size: 2880, dropout=0.8 |
| Output layer | Size: 5, activation primitive: softmax |
The heart beat cycle segmentation data that step 1) obtains are upset into sequence at random, are divided into mutually not phase according to the ratio of about 92:8
The training set (80000) and test set (6383) of friendship, are trained TDCNN using training set, are verified using test set,
The deconditioning when test set precision of prediction reaches setting value (such as 0.95).
As preferred: the step c) carries out intelligent diagnostics to true ECG data using trained network, and looks for
To arrhythmia cordis position, in the heart beat cycle segmentation data input foundation and trained TDCNN by acquisition, each heart is obtained
Hop cycle types of arrhythmia provides arrhythmia cordis position then according to the segmentation data segment where types of arrhythmia.
The invention firstly uses the R wave positions of the MODWT detection practical electrocardiogram of patient, then are based on R wave position and heart rate pair
Data are split, the experimental results showed that, after handling in this way, the accuracy of analysis is greatly improved；Then, by multi-lead
Data carry out two dimensional image processing, are analyzed using TDCNN data, to consider the internal space connection between lead
System.
The present invention can be split the true ECG data of random length according to different people R wave position, heart rate, and fill
Divide the inherent correlation considered between lead, it can be readily appreciated that easy to accomplish, precision of prediction is higher, is more suitable for examining for arrhythmia cordis
It is disconnected.
Detailed description of the invention
Fig. 1 is to detect practical ECG R wave position view using MODWT in the present invention.
Fig. 2 is the one section of practical ECG data divided in the present invention using MODWT.
Specific embodiment
Below in conjunction with attached drawing, the technical scheme of the present invention will be explained in further detail.One kind based on MODWT and
The multi-lead arrhythmia cordis intelligent diagnosing method of TDCNN, comprising the following steps:
A) the true ECG data R wave position of random length is found based on MODWT, data is split；
Equipped with multi-lead ECG data X:
Wherein: m indicates lead number, and n indicates sampling number (because of sample frequency fcIt is known that can obtain at sampled point i to it is corresponding when
Carve ti=i/fc)。
Step 1: taking the 1st leads as analysis data, obtain sampled point locating for continuous s R wave using MODWT programming
Position w1, w2 ..., ws, it is seen that: 1 < w1 < w2 ... < ws < n.
MODWT is the non-orthogonal transformation of high redundancy, and sample size can be arbitrary value, have shift invariant, very suitable
Close processing ECG data.In actual use, wavelet type similar therewith is first chosen according to ECG R wave shape, so
Suitable segmentation level is chosen according to application effect afterwards, detail signal is finally chosen and original signal is reconstructed, find greatly
It is worth point, this is R wave position.
Step 2: obtain the interval RR:
y1=w2-w1,y2=w3-w2,…,ys-1=ws-ws-1
Taking is worth heart rate yz.Because we use MIT-BIH arrhythmia cordis database, selected types of arrhythmia are as follows:
Normal sinus rhythm (N), left bundle branch block (L), right bundle branch block (R), atrial premature beats (A) and ventricular premature beat (V)
Five kinds of types of arrhythmia.In this five seed type, tetra- kinds of N, L, R, V can find out clue by QRS wave, and type-A needs are examined
Consider whether p wave in front occurs in advance, to get p wave, by (the note: generally in heart beat cycle of heart beat cycle position locating for R wave
Front half section) toward moving back, at w1: data intercept are as follows:
Note that when subscript<0 or>n when, give up the segment data.(0,1) normalized is carried out to data, and passes through stretching
Or compressed transform is to fixed length N (note: the numerically equal to sample frequency of training data is 360Hz in embodiment):
x111..., x11N
Wherein, subscript the 1st " 1 " indicates lead 1, the 1st heart beat cycle of the 2nd " 1 " expression, the 1st heart of the 3rd expression
The sampling point position of hop cycle.
Similarly, other positions data intercept is obtained:
x121...,x12N..., x1s1..., x1sN。
Step 3: same to handle on the basis of the 1st lead R wave position to other leads.By each lead in each heart beat cycle
Fixed-length data be sequentially arranged above and below, obtain 2-D data divide and combine Y:
Y=[Y1..., Ys],
Wherein:
This is the 1 heart beat cycle segmentation data that need to be diagnosed.
B) it using MIT-BIH arrhythmia cordis database as training test data source, establishes based on TDCNN and trains nerve
Network
Step 1: MIT-BIH arrhythmia cordis database is obtained, exception lead and the few types of arrhythmia of data volume are given up,
Selected II lead, the normal sinus rhythm (N) of VI lead, left bundle branch block (L), right bundle branch block (R), Fang Xingzao
It fights (A) and five kinds of types of arrhythmia of ventricular premature beat (V) is as analysis data.Because the database has mark to each R wave position
Note, so, R wave position be it is given, be not required to be detected using MODWT, if according to the R wave position of mark, such as a) step
2, the 3 acquisition heart beat cycles divide data；
Step 2: constructing TDCNN according to shown in table 1.
1 TDCNN structure of table
| Input layer | II lead, VI leads dimension: 2 × 360 |
| Convolutional layer 1 | Filter quantity: 32, size: 2 × 9 |
| Pond layer 1 | Size: 2 × 2 |
| Convolutional layer 2 | Filter quantity: 32, size: 1 × 9 |
| Pond layer 2 | Size: 1 × 2 |
| Full articulamentum | Size: 2880, dropout=0.8 |
| Output layer | Quantity: 5, activation primitive: softmax |
It is to be described in detail below:
It is different using one-dimensional CNN from usual processing ECG data, here to consider the space relationship between each lead, adopt
It is handled with TDCNN.For two leads of the MIT-BIH in embodiment, input layer is the sequentially upper and lower of two leads
Fixed-length data is arranged, because single lead sample frequency is 360Hz, so input layer is 2 × 360；Followed by convolutional layer 1, filter
Wave device quantity is 32, single filter be also it is two-dimensional, size is 2 × 9, and step-length takes 1 in both direction, passes through and mends 0 and protect
Length and width size after card convolution is constant, and data dimension is 2 × 360 × 32, and to prevent gradient from disappearing, activation primitive takes ReLu letter
Number；Again by pond layer 1, size is 2 × 2, and step-length takes 2 in both direction, and data dimension becomes 1 × 180 × 32；It connects down
Being convolutional layer 2, filter quantity is 32, and size is 1 × 9, and step-length takes 1 in both direction, data dimension becomes 1 ×
180 × 32, activation primitive still takes ReLu function；Again by pond layer 2, size is 1 × 2, and step-length takes 1 × 2, and data dimension becomes
It is 1 × 90 × 32；It next is full articulamentum, size 2880 is handled by dropout to prevent over-fitting, taken
0.8；It is finally output layer, using Softmax classifier.Training network loss function uses cross entropy, using batch mode,
Batch size is 150, and assessment models use precision of prediction:
Positive exact figures/the sum of precision of prediction=prediction.
The heart beat cycle segmentation data that step 1 obtains are upset into sequence at random, are divided into mutually not phase according to the ratio of about 92:8
The training set (80000) and test set (6383) of friendship, are trained TDCNN using training set, are verified using test set,
The deconditioning when test set precision of prediction reaches setting value (such as 0.95).
C) intelligent diagnostics are carried out to true ECG data using trained network, and finds arrhythmia cordis position.
A) middle acquisition is inputted into neural network to diagnostic signal Y, obtains types of arrhythmia, and provide the time at place
Position is referred to for doctor.
Embodiment: in order to examine diagnostic result, we are using patient ECG's data that a expert marked as follow-up
Break practical ECG signal: lead: II lead, VI lead；Duration: 30 minutes；Sample frequency: 360Hz；In R wave position, have
Mark at 2526: normal sinus rhythm (N) marks at 41: ventricular premature beat (V).
Calculated result: according to abovementioned steps a): it is as shown in Figure 1 to find R wave position using MODWT, it is seen that program looks arrive
R wave position almost with expert mark R wave position it is completely the same.Seeking the interval the RR eartbeat interval that must be averaged is 242 times, i.e. the heart
Rate is about are as follows: 60/ (242/360)=90bpm.360 sampled points are normalized and are stretched over, to two leads being sequentially arranged above and below
Data are split by R wave position and heart rate, obtain 2565 parts of data, we are with the 3rd part of double leads as an example, such as Fig. 2
It is shown.
According to abovementioned steps b): MIT-BIH arrhythmia cordis database is obtained, and is split according to the R wave position of mark,
Because average heart rate is not consistent in every part of record, every part of two leads according to heart rate segmentation are drawn high or are compressed to 360
A sampled point obtains 86383 parts, then upsets sequence at random, is divided into mutually disjoint training set (80000) according to the ratio of about 92:8
With test set (6383), TDCNN is established, then TDCNN is trained using training set, is verified using test set.
It is according to tune ginseng as a result, preferable according to effect when establishing TDCNN shown in table 1, when the number of iterations reaches 2000 times, in advance
It surveys precision and has reached 0.98, meet the expected requirements.
According to abovementioned steps c): the ECG signal to be diagnosed that step a) is generated inputs TDCNN, has obtained arrhythmia cordis
Diagnostic result, precision reach 0.994, and obtain time location where arrhythmia cordis, and the diagnosis for next step doctor provides ginseng
It examines.
The above examples only illustrate the technical idea of the present invention, and this does not limit the scope of protection of the present invention, all
According to the technical idea provided by the invention, any changes made on the basis of the technical scheme each falls within the scope of the present invention
Within.
Claims (3)
1. a kind of multi-lead arrhythmia cordis intelligent diagnosing method based on MODWT and TDCNN, it includes the following steps:
A) the true ECG data R wave position of random length is found based on MODWT, data is split；
B) it using MIT-BIH arrhythmia cordis database as training test data source, establishes based on TDCNN and trains nerve net
Network；
C) intelligent diagnostics are carried out to true ECG data using trained network, and finds arrhythmia cordis position；Its feature
It is:
The step a) finds the true ECG data R wave position of random length based on MODWT, is split method to data
Are as follows: it is equipped with multi-lead ECG data X:
Wherein: m indicates lead number, and n indicates sampling number (because of sample frequency fcMoment t is corresponded at sampled point i it is known that can obtaini
=i/fc), include the following steps:
Step 1: taking the 1st leads as analysis data, obtain sampling point position locating for continuous s R wave using MODWT programming
w1、w2、…、ws, it is seen that: 1 < w1<w2…<ws<n；
Step 2: obtain the interval RR:
y1=w2-w1,y2=w3-w2,…,ys-1=ws-ws-1
Taking is worth heart rate yz, for comprising a heart beat cycle, in w1Place: data intercept are as follows:
Note that when subscript<0 or>n when, give up the segment data.Data are normalized, and are become by stretching or compressing
Change to fixed length N (sample frequency of training data note: can be set as):
x111..., x11N
Wherein, subscript the 1st " 1 " indicates lead 1, the 1st heart beat cycle of the 2nd " 1 " expression, the 3rd the 1st heartbeat week of expression
The sampling point position of phase；Similarly, other positions data intercept is obtained:
x121..., x12N, x131..., x13N..., x1s1..., x1sN。
Step 3: same to handle on the basis of the 1st lead R wave position to other leads；By each lead determining in each heart beat cycle
Long data are sequentially arranged above and below, and obtain the 2-D data combination Y divided:
Y=[Y1..., Ys],
Wherein:
This is the heart beat cycle segmentation data that need to be diagnosed.
2. the multi-lead arrhythmia cordis intelligent diagnosing method according to claim 1 based on MODWT and TDCNN, feature
It is:
The step b) is established and is instructed based on TDCNN using MIT-BIH arrhythmia cordis database as training test data source
Practice neural network method are as follows:
Step 1): obtaining MIT-BIH arrhythmia cordis database, gives up exception lead and the few types of arrhythmia of data volume, choosing
Fixed II lead, the normal sinus rhythm (N) of VI lead, left bundle branch block (L), right bundle branch block (R), atrial premature beats
(A) and five kinds of types of arrhythmia of ventricular premature beat (V) are as analysis data；According to the R wave position of mark, heart beat cycle point is obtained
Cut data；
Step 2): TDCNN is constructed according to such as the following table 1；
1 TDCNN structure of table
The heart beat cycle segmentation data that step 1) obtains are upset into sequence at random, are divided into according to the ratio of about 92:8 mutually disjoint
Training set (80000) and test set (6383), are trained TDCNN using training set, are verified using test set, work as survey
Deconditioning when examination collection precision of prediction reaches setting value (such as 0.95).
3. the multi-lead arrhythmia cordis intelligent diagnosing method according to claim 1 or 2 based on MODWT and TDCNN, special
Sign is:
The step c) carries out intelligent diagnostics to true ECG data using trained network, and finds arrhythmia cordis position
It sets, in the heart beat cycle segmentation data input foundation and trained TDCNN by acquisition, obtains each heart beat cycle rhythm of the heart and lose
Normal type provides arrhythmia cordis position then according to the segmentation data segment where types of arrhythmia.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811544778.3A CN109887594A (en) | 2018-12-17 | 2018-12-17 | A kind of multi-lead arrhythmia cordis intelligent diagnosing method based on MODWT and TDCNN |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811544778.3A CN109887594A (en) | 2018-12-17 | 2018-12-17 | A kind of multi-lead arrhythmia cordis intelligent diagnosing method based on MODWT and TDCNN |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109887594A true CN109887594A (en) | 2019-06-14 |
Family
ID=66925122
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811544778.3A Pending CN109887594A (en) | 2018-12-17 | 2018-12-17 | A kind of multi-lead arrhythmia cordis intelligent diagnosing method based on MODWT and TDCNN |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109887594A (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110141220A (en) * | 2019-06-20 | 2019-08-20 | 鲁东大学 | Myocardial infarction automatic testing method based on multi-modal fusion neural network |
| CN110522444A (en) * | 2019-09-03 | 2019-12-03 | 西安邮电大学 | A kind of electrocardiosignal method for identifying and classifying based on Kernel-CNN |
| CN111543979A (en) * | 2020-05-13 | 2020-08-18 | 许祥林 | Method for outputting vector cardiogram through conventional leads |
| CN112587147A (en) * | 2020-11-11 | 2021-04-02 | 上海数创医疗科技有限公司 | Atrial premature beat target detection method based on convolutional neural network |
| WO2021098461A1 (en) * | 2019-11-22 | 2021-05-27 | 华为技术有限公司 | Heart state evaluation method and apparatus, terminal device and storage medium |
| CN115316996A (en) * | 2021-05-10 | 2022-11-11 | 广州视源电子科技股份有限公司 | Training method, device and equipment for abnormal heart rhythm recognition model and storage medium |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1951320A (en) * | 2006-11-16 | 2007-04-25 | 上海交通大学 | Abnormal electrocardiogram recognition method based on ultra-complete characteristics |
| CN104102915A (en) * | 2014-07-01 | 2014-10-15 | 清华大学深圳研究生院 | Multiple-template matching identity recognition method based on ECG (Electrocardiogram) under electrocardiogram abnormality state |
| US20150164413A1 (en) * | 2013-12-13 | 2015-06-18 | National Chung Shan Institute Of Science And Technology | Method of creating anesthetic consciousness index with artificial neural network |
| CN105748063A (en) * | 2016-04-25 | 2016-07-13 | 山东大学齐鲁医院 | Intelligent arrhythmia diagnosis method based on multiple-lead and convolutional neural network |
| CN107822622A (en) * | 2017-09-22 | 2018-03-23 | 成都比特律动科技有限责任公司 | Electrocardiographic diagnosis method and system based on depth convolutional neural networks |
| CN108932452A (en) * | 2017-05-22 | 2018-12-04 | 中国科学院半导体研究所 | Arrhythmia cordis beat classification method based on multiple dimensioned convolutional neural networks |
-
2018
- 2018-12-17 CN CN201811544778.3A patent/CN109887594A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1951320A (en) * | 2006-11-16 | 2007-04-25 | 上海交通大学 | Abnormal electrocardiogram recognition method based on ultra-complete characteristics |
| US20150164413A1 (en) * | 2013-12-13 | 2015-06-18 | National Chung Shan Institute Of Science And Technology | Method of creating anesthetic consciousness index with artificial neural network |
| CN104102915A (en) * | 2014-07-01 | 2014-10-15 | 清华大学深圳研究生院 | Multiple-template matching identity recognition method based on ECG (Electrocardiogram) under electrocardiogram abnormality state |
| CN105748063A (en) * | 2016-04-25 | 2016-07-13 | 山东大学齐鲁医院 | Intelligent arrhythmia diagnosis method based on multiple-lead and convolutional neural network |
| CN108932452A (en) * | 2017-05-22 | 2018-12-04 | 中国科学院半导体研究所 | Arrhythmia cordis beat classification method based on multiple dimensioned convolutional neural networks |
| CN107822622A (en) * | 2017-09-22 | 2018-03-23 | 成都比特律动科技有限责任公司 | Electrocardiographic diagnosis method and system based on depth convolutional neural networks |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110141220A (en) * | 2019-06-20 | 2019-08-20 | 鲁东大学 | Myocardial infarction automatic testing method based on multi-modal fusion neural network |
| CN110522444A (en) * | 2019-09-03 | 2019-12-03 | 西安邮电大学 | A kind of electrocardiosignal method for identifying and classifying based on Kernel-CNN |
| CN110522444B (en) * | 2019-09-03 | 2022-03-25 | 西安邮电大学 | Electrocardiosignal identification and classification method based on Kernel-CNN |
| WO2021098461A1 (en) * | 2019-11-22 | 2021-05-27 | 华为技术有限公司 | Heart state evaluation method and apparatus, terminal device and storage medium |
| CN111543979A (en) * | 2020-05-13 | 2020-08-18 | 许祥林 | Method for outputting vector cardiogram through conventional leads |
| CN111543979B (en) * | 2020-05-13 | 2024-02-06 | 许祥林 | Method for outputting electrocardiographic vector diagram by conventional leads |
| CN112587147A (en) * | 2020-11-11 | 2021-04-02 | 上海数创医疗科技有限公司 | Atrial premature beat target detection method based on convolutional neural network |
| CN115316996A (en) * | 2021-05-10 | 2022-11-11 | 广州视源电子科技股份有限公司 | Training method, device and equipment for abnormal heart rhythm recognition model and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109887594A (en) | A kind of multi-lead arrhythmia cordis intelligent diagnosing method based on MODWT and TDCNN | |
| CN108714026B (en) | Fine-grained electrocardiosignal classification method based on deep convolutional neural network and online decision fusion | |
| CN111990989A (en) | Electrocardiosignal identification method based on generation countermeasure and convolution cyclic network | |
| CN106108889B (en) | Electrocardiogram classification method based on deep learning algorithm | |
| CN109171712A (en) | Auricular fibrillation recognition methods, device, equipment and computer readable storage medium | |
| CN107822622A (en) | Electrocardiographic diagnosis method and system based on depth convolutional neural networks | |
| CN113397555A (en) | Arrhythmia classification algorithm of C-LSTM for physiological parameter monitoring | |
| CN106901723A (en) | A kind of electrocardiographic abnormality automatic diagnosis method | |
| CN109394205B (en) | Electrocardiosignal analysis method based on deep neural network | |
| CN107811626A (en) | A kind of arrhythmia classification method based on one-dimensional convolutional neural networks and S-transformation | |
| CN106214145A (en) | A kind of electrocardiogram classification method based on degree of depth learning algorithm | |
| CN110638430B (en) | Method for building cascade neural network ECG signal arrhythmia classification model | |
| CN112906748A (en) | 12-lead ECG arrhythmia detection classification model construction method based on residual error network | |
| CN104523266A (en) | Automatic classification method for electrocardiogram signals | |
| CN110236529A (en) | A kind of multi-lead arrhythmia cordis intelligent diagnosing method based on MODWT and LSTM | |
| CN113440149B (en) | ECG signal classification method based on twelve-lead electrocardiograph data two-dimensional multi-input residual neural network | |
| CN111626114A (en) | Electrocardiosignal arrhythmia classification system based on convolutional neural network | |
| Gawande et al. | Heart diseases classification using convolutional neural network | |
| CN109222963A (en) | A kind of anomalous ecg method for identifying and classifying based on convolutional neural networks | |
| Khan et al. | Electrocardiogram heartbeat classification using convolutional neural networks for the detection of cardiac Arrhythmia | |
| CN109620203A (en) | A kind of electrocardiosignal characteristic automatic extraction method based on one-dimensional convolutional neural networks | |
| CN107480637A (en) | Heart failure based on heart sound feature method by stages | |
| CN116361688A (en) | Multi-mode feature fusion model construction method for automatic classification of electrocardiographic rhythms | |
| CN106063704B (en) | A kind of QRS wave starting and terminal point localization method based on regularization least square recurrence learning | |
| CN110537907B (en) | Electrocardiosignal compression and identification method based on singular value decomposition |
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 | ||
| CB02 | Change of applicant information |
Address after: 310012 block B, 5 / F, building e, Paradise Software Park, No.3 xidoumen Road, Xihu District, Hangzhou City, Zhejiang Province Applicant after: ZHEJIANG HELOWIN MEDICAL TECHNOLOGY Co.,Ltd. Address before: 310012 A building 7D, Paradise Software Park, 3 West Road, Hangzhou, Zhejiang, Xihu District Applicant before: ZHEJIANG HELOWIN MEDICAL TECHNOLOGY Co.,Ltd. |
|
| CB02 | Change of applicant information | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190614 |
|
| RJ01 | Rejection of invention patent application after publication |