CN114841209A - Multi-target domain electrocardiosignal classification method based on depth field self-adaption - Google Patents

Multi-target domain electrocardiosignal classification method based on depth field self-adaption Download PDF

Info

Publication number
CN114841209A
CN114841209A CN202210526548.4A CN202210526548A CN114841209A CN 114841209 A CN114841209 A CN 114841209A CN 202210526548 A CN202210526548 A CN 202210526548A CN 114841209 A CN114841209 A CN 114841209A
Authority
CN
China
Prior art keywords
features
layer
branch unit
convolution
lead
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
CN202210526548.4A
Other languages
Chinese (zh)
Inventor
舒明雷
许会芳
田岚
刘辉
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.)
Shandong University
Shandong Institute of Artificial Intelligence
Original Assignee
Shandong University
Shandong Institute of Artificial Intelligence
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 Shandong University, Shandong Institute of Artificial Intelligence filed Critical Shandong University
Priority to CN202210526548.4A priority Critical patent/CN114841209A/en
Publication of CN114841209A publication Critical patent/CN114841209A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/12Classification; Matching
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00Computing arrangements based on biological models
    • G06N3/02Neural networks
    • G06N3/04Architecture, e.g. interconnection topology
    • G06N3/045Combinations of networks
    • 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

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Artificial Intelligence (AREA)
  • General Engineering & Computer Science (AREA)
  • Biomedical Technology (AREA)
  • Molecular Biology (AREA)
  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Software Systems (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • General Health & Medical Sciences (AREA)
  • Mathematical Physics (AREA)
  • Computing Systems (AREA)
  • Signal Processing (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)

Abstract

A multi-target domain electrocardiosignal classification method based on depth field self-adaptation does not need feature alignment between two domains, focuses on a prediction method of the category accuracy and diversity of a target domain, improves the generalization capability of a model on the multi-target domain, has high convergence speed, and solves the problem of poor effect of a single target domain adaptation algorithm on the multi-target domain. In addition, the introduction of the deformable convolution kernel can consciously adjust the field of experience of convolution in the model training process, and better adapt to the lead characteristics of the multi-lead electrocardiosignals.

Description

Multi-target domain electrocardiosignal classification method based on depth field self-adaption
Technical Field
The invention relates to the field of electrocardiosignal classification, in particular to a multi-target domain electrocardiosignal classification method based on depth field self-adaptation.
Background
At present, the existing intelligent electrocardiosignal classification method has good effect in labeling a large amount of electrocardiosignal data, however, in practical application, the labeling of the electrocardiosignal needs to depend on expert knowledge, and the method needs a large amount of manpower and time cost. The development of the field self-adaption in the transfer learning provides a thought for solving the problem, namely, a neural network trained on a source domain data set (labeled) can be highly suitable for a target domain data set (unlabeled) with characteristic distribution obviously different from that of a source domain, and the problem of characteristic distribution misalignment of the source domain and the target domain can be effectively solved.
Disclosure of Invention
In order to overcome the defects of the technology, the invention provides a method for improving the accuracy of the electrocardio classification of a plurality of target domains on a classification model trained by a source domain.
The technical scheme adopted by the invention for overcoming the technical problems is as follows:
a multi-target domain electrocardiosignal classification method based on depth field self-adaptation comprises the following steps:
a) acquiring a plurality of 12-lead electrocardiosignals acquired by different acquisition equipment and different individuals, labeling the 12-lead electrocardiosignals, dividing the labeled 12-lead electrocardiosignals into a source domain S, and dividing the unlabeled 12-lead electrocardiosignals into a target domain combination T, wherein T is ═ T { T } 1 ,T 2 ,...,T k ,...,T M },T k The number of target domains is k, wherein k is 1, 2.
b) Respectively and sequentially performing down-sampling processing, slicing processing and normalization processing on the source domain S and the target domain combination T to obtain a processed source domain S 'and a processed target domain combination T';
c) establishing a deep learning classification model of 12-lead electrocardiosignals, wherein the deep learning classification model of the 12-lead electrocardiosignals is characterized byThe extractor and the full-connection classifier are formed, the processed source domain S 'and the processed target domain combination T' are input into a deep learning classification model of 12-lead electrocardiosignals, and the label prediction value of the source domain is output
Figure BDA0003642722460000021
And label prediction value of target domain
Figure BDA0003642722460000022
d) Calculating a loss function
Figure BDA0003642722460000023
Optimizing a deep learning classification model of the 12-lead electrocardiosignals by an Adam optimization algorithm;
e) and respectively inputting the data of each target domain into the optimized deep learning classification model of the 12-lead electrocardiosignals, outputting to obtain a label predicted value of each target domain, and finishing the classification of the electrocardiosignals of the target domains.
Further, the method for processing the slices in the step b) comprises the following steps: the method comprises the steps of down-sampling all 12-lead electrocardiosignals in a source domain S or a target domain combination T to have the same frequency, randomly intercepting 30S of electrocardiosignals, intercepting 12-lead electrocardiosignals with the data length exceeding 30S, and repeatedly filling 12-lead electrocardiosignals with the data length being less than 30S.
Further, step c) comprises the steps of:
c-1) the feature extractor sequentially comprises a first residual error structure, a second residual error structure, a third residual error structure and a fourth residual error structure;
c-2.1) the first residual error structure is composed of a first branch unit, a second branch unit and a third branch unit, wherein the first branch unit is composed of a first convolution layer, a first batch of normalization layers, a first ReLu activation function layer, a second convolution layer, a second batch of normalization layers and a second ReLu activation function layer in sequence, and the processed source domain S 'and the processed target domain combination T' are respectively input into the first branch unit and then respectively output to obtain the lead internal characteristic of the 12-lead electrocardiosignal
Figure BDA0003642722460000024
And features of
Figure BDA0003642722460000025
c-2.2) the second branch unit of the first residual error structure is composed of a convolution layer, a first batch of normalization layers, a first ReLu activation function layer, a DeformConv variable convolution layer, a second batch of normalization layers and a second ReLu activation function layer in sequence, and the processed source domain S 'and the processed target domain combination T' are respectively input into the second branch unit and then respectively output to obtain the inter-lead characteristics of the 12-lead electrocardiosignals
Figure BDA0003642722460000026
And features of
Figure BDA0003642722460000027
c-2.3) the third branch unit of the first residual error structure is composed of a convolution layer and a Maxpool layer in sequence, and the processed source domain S 'and the processed target domain combination T' are respectively input into the third branch unit and then respectively output to obtain the characteristics of 12-lead electrocardiosignals
Figure BDA0003642722460000028
And features of
Figure BDA0003642722460000029
c-2.4) in-lead characterization
Figure BDA0003642722460000031
And features between leads
Figure BDA0003642722460000032
Inputting the spliced features into a Maxpool layer and outputting the Maxpool layer to obtain the features
Figure BDA0003642722460000033
Inner features of leads
Figure BDA0003642722460000034
And features between leads
Figure BDA0003642722460000035
After feature splicing, inputting the feature into a Maxpool layer of the Maxpool layer and outputting the feature to obtain features
Figure BDA0003642722460000036
c-2.5) characterization of
Figure BDA0003642722460000037
And features of
Figure BDA0003642722460000038
Overlap to form a new feature F 1 S Will be characterized by
Figure BDA0003642722460000039
And features of
Figure BDA00036427224600000310
Performing superposition to form new features
Figure BDA00036427224600000311
c-3.1) the second residual structure is composed of a first branch unit, a second branch unit and a third branch unit, wherein the first branch unit is composed of a first convolution layer, a first batch of normalization layers, a first ReLu activation function layer, a second convolution layer, a second batch of normalization layers and a second ReLu activation function layer in sequence, and the characteristics F are respectively formed by 1 S And features of
Figure BDA00036427224600000312
Respectively output the internal lead characteristics of the obtained 12-lead electrocardiosignals after being input into the first branch unit
Figure BDA00036427224600000313
And features of
Figure BDA00036427224600000314
c-3.2) the second branch unit of the second residual structure sequentially comprises a convolution layer, a first batch of normalization layers, a first ReLu activation function layer, a DeformConv variable convolution layer, a second batch of normalization layers and a second ReLu activation function layer, and respectively forms the characteristic F 1 S And features
Figure BDA00036427224600000315
The inter-lead characteristics of 12-lead electrocardiosignals are respectively output after being input into the second branch unit
Figure BDA00036427224600000316
And features of
Figure BDA00036427224600000317
c-3.3) the third branch unit of the second residual structure is composed of a convolution layer and a Maxpool layer in sequence, and respectively converts the characteristic F 1 S And features
Figure BDA00036427224600000318
Respectively outputting the signals to obtain the characteristics of 12-lead electrocardiosignals after being input into a third branch unit
Figure BDA00036427224600000319
And features of
Figure BDA00036427224600000320
c-3.4) in-lead characterization
Figure BDA00036427224600000321
And features between leads
Figure BDA00036427224600000322
Inputting the spliced features into a Maxpool layer and outputting the Maxpool layer to obtain the features
Figure BDA00036427224600000323
Inner features of leads
Figure BDA00036427224600000324
And features between leads
Figure BDA00036427224600000325
After feature splicing, inputting the feature into a Maxpool layer of the Maxpool layer and outputting the feature to obtain features
Figure BDA00036427224600000326
c-3.5) characterization of
Figure BDA00036427224600000327
And features of
Figure BDA00036427224600000328
Performing superposition to form new features
Figure BDA00036427224600000329
Will be characterized by
Figure BDA00036427224600000330
And features of
Figure BDA00036427224600000331
Performing superposition to form new features
Figure BDA00036427224600000332
c-4.1) the third residual error structure is composed of a first branch unit, a second branch unit and a third branch unit, wherein the first branch unit is composed of a first convolution layer, a first batch of normalization layers, a first ReLu activation function layer, a second convolution layer, a second batch of normalization layers and a second ReLu activation function layer in sequence, and the characteristics of the first branch unit, the first batch of normalization layers and the second batch of normalization layers are respectively characterized
Figure BDA0003642722460000041
And features
Figure BDA0003642722460000042
Respectively output the internal lead characteristics of the obtained 12-lead electrocardiosignals after being input into the first branch unit
Figure BDA0003642722460000043
And features of
Figure BDA0003642722460000044
c-4.2) the second branch unit of the third residual error structure sequentially comprises a convolution layer, a first batch of normalization layers, a first ReLu activation function layer, a DeformConv variable convolution layer, a second batch of normalization layers and a second ReLu activation function layer, and the characteristics of the second branch unit are respectively represented by
Figure BDA0003642722460000045
And features
Figure BDA0003642722460000046
The inter-lead characteristics of 12-lead electrocardiosignals are respectively output after being input into the second branch unit
Figure BDA0003642722460000047
And features of
Figure BDA0003642722460000048
c-4.3) the third branch unit of the third residual structure is composed of a convolution layer and a Maxpool layer in sequence, and features of the convolution layer and the Maxpool layer are respectively
Figure BDA0003642722460000049
And features
Figure BDA00036427224600000410
Respectively outputting the signals to obtain the characteristics of 12-lead electrocardiosignals after being input into a third branch unit
Figure BDA00036427224600000411
And features of
Figure BDA00036427224600000412
c-4.4) in-lead characterization
Figure BDA00036427224600000413
And features between leads
Figure BDA00036427224600000414
After feature splicing, inputting the feature into a Maxpool layer of the Maxpool layer and outputting the feature to obtain features
Figure BDA00036427224600000415
Inner features of leads
Figure BDA00036427224600000416
And features between leads
Figure BDA00036427224600000417
Inputting the spliced features into a Maxpool layer and outputting the Maxpool layer to obtain the features
Figure BDA00036427224600000418
c-4.5) characterization of
Figure BDA00036427224600000419
And features of
Figure BDA00036427224600000420
Overlap to form a new feature F 3 S Will be characterized by
Figure BDA00036427224600000421
And features of
Figure BDA00036427224600000422
Performing superposition to form new features
Figure BDA00036427224600000423
c-5.1) the fourth residual structure is composed of a first branch unit, a second branch unit, and a third branch unit,the first branch unit consists of a first convolution layer, a first batch of normalization layers, a first ReLu activation function layer, a second convolution layer, a second batch of normalization layers and a second ReLu activation function layer in sequence, and the first branch unit respectively consists of a characteristic F 3 S And features
Figure BDA00036427224600000424
Respectively output the internal lead characteristics of the obtained 12-lead electrocardiosignals after being input into the first branch unit
Figure BDA00036427224600000425
And features of
Figure BDA00036427224600000426
c-5.2) the second branch unit of the fourth residual structure sequentially comprises a convolution layer, a first batch of normalization layers, a first ReLu activation function layer, a DeformConv variable convolution layer, a second batch of normalization layers and a second ReLu activation function layer, and respectively forms the characteristic F 3 S And features
Figure BDA0003642722460000051
The inter-lead characteristics of 12-lead electrocardiosignals are respectively output after being input into the second branch unit
Figure BDA0003642722460000052
And features of
Figure BDA0003642722460000053
c-5.3) the third branch unit of the fourth residual structure is composed of a convolution layer and a Maxpool layer in sequence, and the characteristics F are respectively obtained 3 S And features
Figure BDA0003642722460000054
Respectively outputting the signals to obtain the characteristics of 12-lead electrocardiosignals after being input into a third branch unit
Figure BDA0003642722460000055
And features of
Figure BDA0003642722460000056
c-5.4) in-lead characterization
Figure BDA0003642722460000057
And features between leads
Figure BDA0003642722460000058
Inputting the spliced features into a Maxpool layer and outputting the Maxpool layer to obtain the features
Figure BDA0003642722460000059
Inner features of leads
Figure BDA00036427224600000510
And features between leads
Figure BDA00036427224600000511
After feature splicing, inputting the feature into a Maxpool layer of the Maxpool layer and outputting the feature to obtain features
Figure BDA00036427224600000512
c-5.5) characterization of
Figure BDA00036427224600000513
And features of
Figure BDA00036427224600000514
Performing superposition to form new features
Figure BDA00036427224600000515
Will be characterized by
Figure BDA00036427224600000516
And features of
Figure BDA00036427224600000517
Performing superposition to form new features
Figure BDA00036427224600000518
c-6) characterization of
Figure BDA00036427224600000519
And features of
Figure BDA00036427224600000520
Inputting the data into a full-connection classifier, and outputting the data to obtain a label predicted value of a source domain
Figure BDA00036427224600000521
And label prediction value of target domain
Figure BDA00036427224600000522
Further, in the step c-2.1), the number of channels of the first convolution layer and the second convolution layer of the first branch unit is 32, the size of the convolution kernel is 1 × 3, in the step c-2.2), the number of channels of the convolution layer of the second branch unit is 32, the size of the convolution kernel is 1 × 3, the number of channels of the DeformConv variable convolution layer is 32, the size of the convolution kernel is 3 × 3, in the step c-2.3), the size of the convolution kernel of the max pool layer is 1 × 4, and in the step c-2.4), the size of the convolution kernel of the max pool layer is 1 × 4.
Further, in step c-3.1), the number of channels of the first convolution layer and the second convolution layer of the first branch unit is 64, the size of the convolution kernel is 1 × 3, in step c-3.2), the number of channels of the convolution layer of the second branch unit is 64, the size of the convolution kernel is 1 × 3, the number of channels of the DeformConv variable convolution layer is 64, the size of the convolution kernel is 3 × 3, in step c-3.3), the size of the convolution kernel of the max pool layer is 1 × 4, and in step c-3.4), the size of the convolution kernel of the max pool layer is 1 × 4.
Further, in step c-4.1), the number of channels of the first convolution layer and the second convolution layer of the first branch unit is 128, the size of the convolution kernel is 1 × 3, in step c-4.2), the number of channels of the convolution layer of the second branch unit is 128, the size of the convolution kernel is 1 × 3, the number of channels of the DeformConv variable convolution layer is 128, the size of the convolution kernel is 3 × 3, in step c-4.3), the size of the convolution kernel of the Maxpool maximum pooling layer is 1 × 4, and in step c-4.4), the size of the convolution kernel of the Maxpool maximum pooling layer is 1 × 4.
Further, in step c-5.1), the number of channels of the first convolution layer and the second convolution layer of the first branch unit is 256, the size of the convolution kernel is 1 × 3, in step c-5.2), the number of channels of the convolution layer of the second branch unit is 256, the size of the convolution kernel is 1 × 3, the number of channels of the DeformConv variable convolution layer is 256, the size of the convolution kernel is 3 × 3, in step c-5.3), the size of the convolution kernel of the Maxpool layer is 1 × 4, and in step c-5.3), the number of channels of the first convolution layer and the number of channels of the second convolution layer are 256, the size of the convolution kernel is 1 × 3
c-5.4) Maxpool layer has a convolution kernel size of 1 × 4.
Further, step d) comprises the following steps:
d-1) by the formula
Figure BDA0003642722460000061
Calculating to obtain cross entropy classification loss
Figure BDA0003642722460000062
In the formula
Figure BDA0003642722460000063
N S Is the total number of samples in the source domain S,
Figure BDA0003642722460000064
is the number of samples with class j in the source domain S, j is 1,2 b In order to train the batch size value,
Figure BDA0003642722460000065
for the true tag value of the ith sample in the source domain S,
Figure BDA0003642722460000066
the label prediction value of the ith sample in the source domain S is obtained;
d-2) by the formula
Figure BDA0003642722460000067
Computationally derived class certainty loss
Figure BDA0003642722460000068
Where C is the number of all classes, ω 1 And omega 2 Are sample deterministic weighting coefficients, T is a transpose,
Figure BDA0003642722460000069
tag prediction values for target domains
Figure BDA00036427224600000610
Prediction probability matrix obtained by Softmax
Figure BDA00036427224600000611
The (c) th column of (a),
Figure BDA00036427224600000612
tag prediction values for target domains
Figure BDA00036427224600000613
Prediction probability matrix obtained by Softmax
Figure BDA00036427224600000614
Column j';
d-3) by the formula
Figure BDA0003642722460000071
Calculating to obtain class diversity loss
Figure BDA0003642722460000072
Figure BDA0003642722460000073
To predict a probability matrix
Figure BDA0003642722460000074
The rank of (d);
d-4) by the formula
Figure BDA0003642722460000075
Calculating a loss function
Figure BDA0003642722460000076
Where α is a coefficient and β is a coefficient.
The invention has the beneficial effects that: the generalization capability of the model on multiple target domains is improved, the convergence speed is high, and the problem that the effect of a single target domain adaptive algorithm on the multiple target domains is poor is solved. In addition, the introduction of the deformable convolution kernel can consciously adjust the field of experience of convolution in the model training process, and better adapt to the lead characteristics of the multi-lead electrocardiosignals.
Drawings
Fig. 1 is a schematic diagram of a network structure according to the present invention.
Detailed Description
The invention is further described below with reference to fig. 1.
A multi-target domain electrocardiosignal classification method based on depth field self-adaptation comprises the following steps:
a) acquiring a plurality of 12-lead electrocardiosignals acquired by different acquisition equipment and different individuals, labeling the 12-lead electrocardiosignals, dividing the labeled 12-lead electrocardiosignals into a source domain S, and dividing the unlabeled 12-lead electrocardiosignals into a target domain combination T, wherein T is ═ T { T } 1 ,T 2 ,...,T k ,...,T M },T k For the kth target field, k is 1, 2.
b) And respectively and sequentially performing down-sampling processing, slicing processing and normalization processing on the source domain S and the target domain combination T to obtain a processed source domain S 'and a processed target domain combination T'.
c) Establishing a deep learning classification model of 12-lead electrocardiosignals, wherein the deep learning classification model of the 12-lead electrocardiosignals consists of a feature extractor and a full-connection classifier, inputting the processed source domain S 'and the processed target domain combination T' into the deep learning classification model of the 12-lead electrocardiosignals, and outputting to obtain a label prediction value of the source domain
Figure BDA0003642722460000077
And label prediction of target domainsValue of
Figure BDA0003642722460000078
d) Calculating a loss function
Figure BDA0003642722460000081
And optimizing a deep learning classification model of the 12-lead electrocardiosignals by an Adam optimization algorithm.
e) And (3) respectively inputting the data of each target domain into the optimized deep learning classification model of the 12-lead electrocardiosignals, outputting to obtain a label predicted value of each target domain, and completing classification of the electrocardiosignals of the target domains.
The method is non-antagonistic, does not need characteristic alignment between two domains, focuses on the prediction method of the category accuracy and diversity of the target domain, improves the generalization capability of the model on the multi-target domain, has high convergence rate, and solves the problem of poor effect of a single target domain adaptive algorithm on the multi-target domain. In addition, the introduction of the deformable convolution kernel can consciously adjust the field of experience of convolution in the model training process, and better adapt to the lead characteristics of the multi-lead electrocardiosignals. The method is not only oriented to a single target domain, but also highly applicable to a plurality of target domains, improves the characteristic distribution difference phenomenon existing among different domain samples, and solves the problem of difficult labeling of the electrical data of a center in practical application.
Example 1:
further, the method for processing the slices in the step b) comprises the following steps: the method comprises the steps of down-sampling all 12-lead electrocardiosignals in a source domain S or a target domain combination T to have the same frequency, randomly intercepting 30S of electrocardiosignals, intercepting 12-lead electrocardiosignals with the data length exceeding 30S, and repeatedly filling 12-lead electrocardiosignals with the data length being less than 30S.
Example 2:
further, step c) comprises the following steps:
c-1) the feature extractor sequentially comprises a first residual error structure, a second residual error structure, a third residual error structure and a fourth residual error structure.
c-2.1) the first residual error structure is composed of a first branch unit, a second branch unit and a third branch unit, wherein the first branch unit is composed of a first convolution layer, a first batch of normalization layers, a first ReLu activation function layer, a second convolution layer, a second batch of normalization layers and a second ReLu activation function layer in sequence, and the processed source domain S 'and the processed target domain combination T' are respectively input into the first branch unit and then respectively output to obtain the lead internal characteristic of the 12-lead electrocardiosignal
Figure BDA0003642722460000082
And features of
Figure BDA0003642722460000083
c-2.2) the second branch unit of the first residual error structure is composed of a convolution layer, a first batch of normalization layers, a first ReLu activation function layer, a DeformConv variable convolution layer, a second batch of normalization layers and a second ReLu activation function layer in sequence, and the processed source domain S 'and the processed target domain combination T' are respectively input into the second branch unit and then respectively output to obtain the inter-lead characteristics of the 12-lead electrocardiosignals
Figure BDA0003642722460000091
And features of
Figure BDA0003642722460000092
c-2.3) the third branch unit of the first residual error structure is composed of a convolution layer and a Maxpool layer in sequence, and the processed source domain S 'and the processed target domain combination T' are respectively input into the third branch unit and then respectively output to obtain the characteristics of 12-lead electrocardiosignals
Figure BDA0003642722460000093
And features of
Figure BDA0003642722460000094
c-2.4) in-lead characterization
Figure BDA0003642722460000095
And features between leads
Figure BDA0003642722460000096
Inputting the spliced features into a Maxpool layer and outputting the Maxpool layer to obtain the features
Figure BDA0003642722460000097
Inner features of leads
Figure BDA0003642722460000098
And features between leads
Figure BDA0003642722460000099
After feature splicing, inputting the feature into a Maxpool layer of the Maxpool layer and outputting the feature to obtain features
Figure BDA00036427224600000910
c-2.5) characterization of
Figure BDA00036427224600000911
And features of
Figure BDA00036427224600000912
Overlap to form a new feature F 1 S Will be characterized by
Figure BDA00036427224600000913
And features of
Figure BDA00036427224600000914
Performing superposition to form new features
Figure BDA00036427224600000915
c-3.1) second residual error StructureThe first branch unit consists of a first convolution layer, a first batch of normalization layers, a first ReLu activation function layer, a second convolution layer, a second batch of normalization layers and a second ReLu activation function layer in sequence, and respectively forms a characteristic F 1 S And features of
Figure BDA00036427224600000916
Respectively output the internal lead characteristics of the obtained 12-lead electrocardiosignals after being input into the first branch unit
Figure BDA00036427224600000917
And features of
Figure BDA00036427224600000918
c-3.2) the second branch unit of the second residual structure sequentially comprises a convolution layer, a first batch of normalization layers, a first ReLu activation function layer, a DeformConv variable convolution layer, a second batch of normalization layers and a second ReLu activation function layer, and respectively forms the characteristic F 1 S And features
Figure BDA00036427224600000919
The inter-lead characteristics of 12-lead electrocardiosignals are respectively output after being input into the second branch unit
Figure BDA00036427224600000920
And features of
Figure BDA00036427224600000921
c-3.3) the third branch unit of the second residual structure is composed of a convolution layer and a Maxpool layer in sequence, and respectively converts the characteristic F 1 S And features
Figure BDA00036427224600000922
Respectively outputting the signals to obtain the characteristics of 12-lead electrocardiosignals after being input into a third branch unit
Figure BDA0003642722460000101
And features of
Figure BDA0003642722460000102
c-3.4) in-lead characterization
Figure BDA0003642722460000103
And inter-lead characteristics
Figure BDA0003642722460000104
Inputting the spliced features into a Maxpool layer and outputting the Maxpool layer to obtain the features
Figure BDA0003642722460000105
Inner features of leads
Figure BDA0003642722460000106
And features between leads
Figure BDA0003642722460000107
Inputting the spliced features into a Maxpool layer with the Maxpool to output the spliced features to obtain the features
Figure BDA0003642722460000108
c-3.5) characterization of
Figure BDA0003642722460000109
And features of
Figure BDA00036427224600001010
Performing superposition to form new features
Figure BDA00036427224600001011
Will be characterized by
Figure BDA00036427224600001012
And features of
Figure BDA00036427224600001013
Are superposed to form a newFeature(s)
Figure BDA00036427224600001014
c-4.1) the third residual error structure is composed of a first branch unit, a second branch unit and a third branch unit, wherein the first branch unit is composed of a first convolution layer, a first batch of normalization layers, a first ReLu activation function layer, a second convolution layer, a second batch of normalization layers and a second ReLu activation function layer in sequence, and the characteristics of the first branch unit, the first batch of normalization layers and the second batch of normalization layers are respectively characterized
Figure BDA00036427224600001015
And features
Figure BDA00036427224600001016
Respectively output the internal lead characteristics of the obtained 12-lead electrocardiosignals after being input into the first branch unit
Figure BDA00036427224600001017
And features of
Figure BDA00036427224600001018
c-4.2) the second branch unit of the third residual error structure sequentially comprises a convolution layer, a first batch of normalization layers, a first ReLu activation function layer, a DeformConv variable convolution layer, a second batch of normalization layers and a second ReLu activation function layer, and the characteristics of the second branch unit are respectively represented by
Figure BDA00036427224600001019
And features
Figure BDA00036427224600001020
The inter-lead characteristics of 12-lead electrocardiosignals are respectively output after being input into the second branch unit
Figure BDA00036427224600001021
And features of
Figure BDA00036427224600001022
c-4.3) second of the third residual StructureThe three-branch unit is composed of a convolution layer and a Maxpool layer in sequence, and features of the three-branch unit are respectively
Figure BDA00036427224600001023
And features
Figure BDA00036427224600001024
Respectively outputting the signals to obtain the characteristics of 12-lead electrocardiosignals after being input into a third branch unit
Figure BDA00036427224600001025
And features of
Figure BDA00036427224600001026
c-4.4) in-lead characterization
Figure BDA00036427224600001027
And features between leads
Figure BDA00036427224600001028
After feature splicing, inputting the feature into a Maxpool layer of the Maxpool layer and outputting the feature to obtain features
Figure BDA00036427224600001029
Inner features of leads
Figure BDA00036427224600001030
And features between leads
Figure BDA00036427224600001031
Inputting the spliced features into a Maxpool layer and outputting the Maxpool layer to obtain the features
Figure BDA00036427224600001032
c-4.5 characterization of
Figure BDA00036427224600001033
And features of
Figure BDA00036427224600001034
Overlap to form a new feature F 3 S Will be characterized by
Figure BDA00036427224600001035
And features of
Figure BDA00036427224600001036
Performing superposition to form new features
Figure BDA00036427224600001037
c-5.1) the fourth residual structure is composed of a first branch unit, a second branch unit and a third branch unit, wherein the first branch unit is composed of a first convolution layer, a first batch of normalization layers, a first ReLu activation function layer, a second convolution layer, a second batch of normalization layers and a second ReLu activation function layer in sequence, and the characteristics F are respectively formed by 3 S And features
Figure BDA0003642722460000111
Respectively output the internal lead characteristics of the obtained 12-lead electrocardiosignals after being input into the first branch unit
Figure BDA0003642722460000112
And features of
Figure BDA0003642722460000113
c-5.2) the second branch unit of the fourth residual structure sequentially comprises a convolution layer, a first batch of normalization layers, a first ReLu activation function layer, a DeformConv variable convolution layer, a second batch of normalization layers and a second ReLu activation function layer, and respectively forms the characteristic F 3 S And features
Figure BDA0003642722460000114
The inter-lead characteristics of 12-lead electrocardiosignals are respectively output after being input into the second branch unit
Figure BDA0003642722460000115
And is characterized bySign for
Figure BDA0003642722460000116
c-5.3) the third branch unit of the fourth residual structure is composed of a convolution layer and a Maxpool layer in sequence, and the characteristics F are respectively obtained 3 S And features
Figure BDA0003642722460000117
Respectively outputting the signals to obtain the characteristics of 12-lead electrocardiosignals after being input into a third branch unit
Figure BDA0003642722460000118
And features of
Figure BDA0003642722460000119
c-5.4) in-lead characterization
Figure BDA00036427224600001110
And features between leads
Figure BDA00036427224600001111
After feature splicing, inputting the feature into a Maxpool layer of the Maxpool layer and outputting the feature to obtain features
Figure BDA00036427224600001112
Inner features of leads
Figure BDA00036427224600001113
And features between leads
Figure BDA00036427224600001114
Inputting the spliced features into a Maxpool layer and outputting the Maxpool layer to obtain the features
Figure BDA00036427224600001115
c-5.5) characterization of
Figure BDA00036427224600001116
And features of
Figure BDA00036427224600001117
Performing superposition to form new features
Figure BDA00036427224600001118
Will be characterized by
Figure BDA00036427224600001119
And features of
Figure BDA00036427224600001120
Performing superposition to form new features
Figure BDA00036427224600001121
c-6) characterization of
Figure BDA00036427224600001122
And features of
Figure BDA00036427224600001123
Inputting the data into a full-connection classifier, and outputting the data to obtain a label predicted value of a source domain
Figure BDA00036427224600001124
And label prediction value of target domain
Figure BDA00036427224600001125
Example 3:
further, in the step c-2.1), the number of channels of the first convolution layer and the second convolution layer of the first branch unit is 32, the size of the convolution kernel is 1 × 3, in the step c-2.2), the number of channels of the convolution layer of the second branch unit is 32, the size of the convolution kernel is 1 × 3, the number of channels of the DeformConv variable convolution layer is 32, the size of the convolution kernel is 3 × 3, in the step c-2.3), the size of the convolution kernel of the Maxpool maximum pooling layer is 1 × 4, and in the step c-2.4), the size of the convolution kernel of the Maxpool maximum pooling layer is 1 × 4.
Example 4:
further, in step c-3.1), the number of channels of the first convolution layer and the second convolution layer of the first branch unit is 64, the size of the convolution kernel is 1 × 3, in step c-3.2), the number of channels of the convolution layer of the second branch unit is 64, the size of the convolution kernel is 1 × 3, the number of channels of the DeformConv variable convolution layer is 64, the size of the convolution kernel is 3 × 3, in step c-3.3), the size of the convolution kernel of the max pool layer is 1 × 4, and in step c-3.4), the size of the convolution kernel of the max pool layer is 1 × 4.
Example 5:
further, in step c-4.1), the number of channels of the first convolution layer and the second convolution layer of the first branch unit is 128, the size of the convolution kernel is 1 × 3, in step c-4.2), the number of channels of the convolution layer of the second branch unit is 128, the size of the convolution kernel is 1 × 3, the number of channels of the DeformConv variable convolution layer is 128, the size of the convolution kernel is 3 × 3, in step c-4.3), the size of the convolution kernel of the Maxpool maximum pooling layer is 1 × 4, and in step c-4.4), the size of the convolution kernel of the Maxpool maximum pooling layer is 1 × 4.
Example 6:
further, in step c-5.1), the number of channels of the first convolution layer and the second convolution layer of the first branch unit is 256, the size of the convolution kernel is 1 × 3, in step c-5.2), the number of channels of the convolution layer of the second branch unit is 256, the size of the convolution kernel is 1 × 3, the number of channels of the DeformConv variable convolution layer is 256, the size of the convolution kernel is 3 × 3, in step c-5.3), the size of the convolution kernel of the max pool layer is 1 × 4, and in step c-5.4), the size of the convolution kernel of the max pool layer is 1 × 4.
Example 7:
further, step d) comprises the following steps:
the training process is divided into three branches, the first branch is supervised training of the source domain data of the label, and the label prediction value of the source domain data
Figure BDA0003642722460000121
And label prediction value of target domain
Figure BDA0003642722460000122
Computing cross entropy classification loss
Figure BDA0003642722460000123
The second branch is a pseudo label prediction matrix for each target domain data
Figure BDA0003642722460000124
Class-deterministic loss function forming a target domain
Figure BDA0003642722460000125
The third branch is a pseudo label prediction matrix of the target domain data
Figure BDA0003642722460000131
Constructed class diversity loss function
Figure BDA0003642722460000132
The three components form a combined loss
Figure BDA0003642722460000133
The parameters of the network model are optimized through an Adam optimization algorithm, so that the joint loss is minimum, the model has good classification performance, the class prediction accuracy of a target domain test set is better, and the model trained by the target domain in a source domain is highly applicable. Specifically, the method comprises the following steps:
d-1) by the formula
Figure BDA0003642722460000134
Calculating to obtain cross entropy classification loss
Figure BDA0003642722460000135
In the formula
Figure BDA0003642722460000136
N S Is the total number of samples in the source domain S,
Figure BDA0003642722460000137
the number of samples of the source domain S with the category j, j 1,2Amount, N b In order to train the batch size value,
Figure BDA0003642722460000138
for the true tag value of the ith sample in the source domain S,
Figure BDA0003642722460000139
and predicting the label of the ith sample in the source domain S.
d-2) by the formula
Figure BDA00036427224600001310
Computationally derived class certainty loss
Figure BDA00036427224600001311
Where C is the number of all classes, ω 1 And omega 2 Are sample deterministic weighting coefficients, T is a transpose,
Figure BDA00036427224600001312
tag prediction values for target domains
Figure BDA00036427224600001313
Prediction probability matrix obtained by Softmax
Figure BDA00036427224600001314
The (c) th column of (a),
Figure BDA00036427224600001315
tag prediction values for target domains
Figure BDA00036427224600001316
Prediction probability matrix obtained by Softmax
Figure BDA00036427224600001317
Column j' of (1).
d-3) by the formula
Figure BDA00036427224600001318
Calculating to obtain category diversityLoss of power
Figure BDA00036427224600001319
Figure BDA00036427224600001320
To predict a probability matrix
Figure BDA00036427224600001321
Is determined.
d-4) by the formula
Figure BDA0003642722460000141
Calculating a loss function
Figure BDA0003642722460000142
Where α is a coefficient and β is a coefficient.
Finally, it should be noted that: although the present invention has been described in detail with reference to the foregoing embodiments, it will be apparent to those skilled in the art that changes may be made in the embodiments and/or equivalents thereof without departing from the spirit and scope of the invention. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims (8)

1. A multi-target domain electrocardiosignal classification method based on depth field self-adaptation is characterized by comprising the following steps:
a) acquiring a plurality of 12-lead electrocardiosignals acquired by different acquisition equipment and different individuals, labeling the 12-lead electrocardiosignals, dividing the labeled 12-lead electrocardiosignals into a source domain S, and dividing the unlabeled 12-lead electrocardiosignals into a target domain combination T, wherein T is ═ T { T } 1 ,T 2 ,...,T k ,...,T M },T k The number of target domains is k, wherein k is 1, 2.
b) Respectively and sequentially performing down-sampling processing, slicing processing and normalization processing on the source domain S and the target domain combination T to obtain a processed source domain S 'and a processed target domain combination T';
c) establishing a deep learning classification model of 12-lead electrocardiosignals, wherein the deep learning classification model of the 12-lead electrocardiosignals consists of a feature extractor and a full-connection classifier, inputting the processed source domain S 'and the processed target domain combination T' into the deep learning classification model of the 12-lead electrocardiosignals, and outputting to obtain a label prediction value of the source domain
Figure FDA0003642722450000011
And label prediction value of target domain
Figure FDA0003642722450000012
d) Calculating a loss function
Figure FDA0003642722450000013
Optimizing a deep learning classification model of the 12-lead electrocardiosignal by an Adam optimization algorithm;
e) and (3) respectively inputting the data of each target domain into the optimized deep learning classification model of the 12-lead electrocardiosignals, outputting to obtain a label predicted value of each target domain, and completing classification of the electrocardiosignals of the target domains.
2. The multi-target domain electrocardiosignal classification method based on the depth domain self-adaptation as claimed in claim 1, wherein the slicing processing method in step b) is as follows: the method comprises the steps of down-sampling all 12-lead electrocardiosignals in a source domain S or a target domain combination T to have the same frequency, randomly intercepting 30S of electrocardiosignals, intercepting 12-lead electrocardiosignals with the data length exceeding 30S, and repeatedly filling 12-lead electrocardiosignals with the data length being less than 30S.
3. The method for classifying multi-target domain electrocardiosignals based on depth domain self-adaptation according to claim 1, wherein the step c) comprises the following steps:
c-1) the feature extractor sequentially comprises a first residual error structure, a second residual error structure, a third residual error structure and a fourth residual error structure;
c-2.1) the first residual error structure is composed of a first branch unit, a second branch unit and a third branch unit, wherein the first branch unit is composed of a first convolution layer, a first batch of normalization layers, a first ReLu activation function layer, a second convolution layer, a second batch of normalization layers and a second ReLu activation function layer in sequence, and the processed source domain S 'and the processed target domain combination T' are respectively input into the first branch unit and then respectively output to obtain the lead internal characteristic of the 12-lead electrocardiosignal
Figure FDA0003642722450000021
And features of
Figure FDA0003642722450000022
c-2.2) the second branch unit of the first residual error structure is composed of a convolution layer, a first batch of normalization layers, a first ReLu activation function layer, a DeformConv variable convolution layer, a second batch of normalization layers and a second ReLu activation function layer in sequence, and the processed source domain S 'and the processed target domain combination T' are respectively input into the second branch unit and then respectively output to obtain the inter-lead characteristics of the 12-lead electrocardiosignals
Figure FDA0003642722450000023
And features of
Figure FDA0003642722450000024
c-2.3) the third branch unit of the first residual error structure is composed of a convolution layer and a Maxpool layer in sequence, and the processed source domain S 'and the processed target domain combination T' are respectively input into the third branch unit and then respectively output to obtain the characteristics of 12-lead electrocardiosignals
Figure FDA0003642722450000025
And features of
Figure FDA0003642722450000026
c-2.4) in-lead characterization
Figure FDA0003642722450000027
And features between leads
Figure FDA0003642722450000028
Inputting the spliced features into a Maxpool layer and outputting the Maxpool layer to obtain the features
Figure FDA0003642722450000029
Inner features of leads
Figure FDA00036427224500000210
And features between leads
Figure FDA00036427224500000211
Inputting the spliced features into a Maxpool layer with the Maxpool to output the spliced features to obtain the features
Figure FDA00036427224500000212
c-2.5) characterization of
Figure FDA00036427224500000213
And features of
Figure FDA00036427224500000214
Overlap to form a new feature F 1 S Will be characterized by
Figure FDA00036427224500000215
And features of
Figure FDA00036427224500000216
Performing superposition to form new features
Figure FDA00036427224500000217
c-3.1) the second residual structure is composed of a first branch unit, a second branch unit and a third branch unit, wherein the first branch unit is composed of a first convolution layer, a first batch of normalization layers, a first ReLu activation function layer, a second convolution layer, a second batch of normalization layers and a second ReLu activation function layer in sequence, and the characteristics F are respectively formed by 1 S And features
Figure FDA00036427224500000218
Respectively output the internal lead characteristics of the obtained 12-lead electrocardiosignals after being input into the first branch unit
Figure FDA00036427224500000219
And features of
Figure FDA00036427224500000220
c-3.2) the second branch unit of the second residual structure sequentially comprises a convolution layer, a first batch of normalization layers, a first ReLu activation function layer, a DeformConv variable convolution layer, a second batch of normalization layers and a second ReLu activation function layer, and respectively forms the characteristic F 1 S And features
Figure FDA0003642722450000031
The inter-lead characteristics of 12-lead electrocardiosignals are respectively output after being input into the second branch unit
Figure FDA0003642722450000032
And features of
Figure FDA0003642722450000033
c-3.3) the third branch unit of the second residual structure is composed of a convolution layer and a Maxpool layer in sequence, and respectively converts the characteristic F 1 S And features
Figure FDA0003642722450000034
Respectively outputting the signals to obtain the characteristics of 12-lead electrocardiosignals after being input into a third branch unit
Figure FDA0003642722450000035
And features of
Figure FDA0003642722450000036
c-3.4) in-lead characterization
Figure FDA0003642722450000037
And features between leads
Figure FDA0003642722450000038
Inputting the spliced features into a Maxpool layer to output the obtained features
Figure FDA0003642722450000039
Inner features of leads
Figure FDA00036427224500000310
And features between leads
Figure FDA00036427224500000311
After feature splicing, inputting the feature into a Maxpool layer of the Maxpool layer and outputting the feature to obtain features
Figure FDA00036427224500000312
c-3.5) characterization of
Figure FDA00036427224500000313
And features of
Figure FDA00036427224500000314
Performing superposition to form new features
Figure FDA00036427224500000315
Will be characterized by
Figure FDA00036427224500000316
And features of
Figure FDA00036427224500000317
Performing superposition to form new features
Figure FDA00036427224500000318
c-4.1) the third residual error structure is composed of a first branch unit, a second branch unit and a third branch unit, wherein the first branch unit is composed of a first convolution layer, a first batch of normalization layers, a first ReLu activation function layer, a second convolution layer, a second batch of normalization layers and a second ReLu activation function layer in sequence, and the characteristics of the first branch unit, the first batch of normalization layers and the second batch of normalization layers are respectively characterized
Figure FDA00036427224500000319
And features
Figure FDA00036427224500000320
Respectively output the internal lead characteristics of the obtained 12-lead electrocardiosignals after being input into the first branch unit
Figure FDA00036427224500000321
And features of
Figure FDA00036427224500000322
c-4.2) the second branch unit of the third residual error structure sequentially comprises a convolution layer, a first batch of normalization layers, a first ReLu activation function layer, a DeformConv variable convolution layer, a second batch of normalization layers and a second ReLu activation function layer, and the characteristics of the second branch unit are respectively represented by
Figure FDA00036427224500000323
And features
Figure FDA00036427224500000324
The inter-lead characteristics of 12-lead electrocardiosignals are respectively output after being input into the second branch unit
Figure FDA00036427224500000325
And features of
Figure FDA00036427224500000326
c-4.3) the third branch unit of the third residual structure is composed of a convolution layer and a Maxpool layer in sequence, and features of the convolution layer and the Maxpool layer are respectively
Figure FDA00036427224500000327
And features
Figure FDA00036427224500000328
Respectively outputting the signals to obtain the characteristics of 12-lead electrocardiosignals after being input into a third branch unit
Figure FDA00036427224500000329
And features of
Figure FDA00036427224500000330
c-4.4) in-lead characterization
Figure FDA00036427224500000331
And features between leads
Figure FDA00036427224500000332
Inputting the spliced features into a Maxpool layer and outputting the Maxpool layer to obtain the features
Figure FDA0003642722450000041
Inner features of leads
Figure FDA0003642722450000042
And features between leads
Figure FDA0003642722450000043
After feature splicing, inputting the feature into a Maxpool layer of the Maxpool layer and outputting the feature to obtain features
Figure FDA0003642722450000044
c-4.5) characterization of
Figure FDA0003642722450000045
And features of
Figure FDA0003642722450000046
Performing superposition to form new features
Figure FDA0003642722450000047
Will be characterized by
Figure FDA0003642722450000048
And features of
Figure FDA0003642722450000049
Performing superposition to form new features
Figure FDA00036427224500000410
c-5.1) the fourth residual structure is composed of a first branch unit, a second branch unit and a third branch unit, wherein the first branch unit is composed of a first convolution layer, a first batch of normalization layers, a first ReLu activation function layer, a second convolution layer, a second batch of normalization layers and a second ReLu activation function layer in sequence, and the characteristics are respectively formed
Figure FDA00036427224500000411
And features
Figure FDA00036427224500000412
Respectively output the internal lead characteristics of the obtained 12-lead electrocardiosignals after being input into the first branch unit
Figure FDA00036427224500000413
And features of
Figure FDA00036427224500000414
c-5.2) the second branch unit of the fourth residual structure sequentially comprises a convolution layer, a first batch of normalization layers, a first ReLu activation function layer, a DeformConv variable convolution layer, a second batch of normalization layers and a second ReLu activation function layer, and the characteristics are respectively formed
Figure FDA00036427224500000415
And features
Figure FDA00036427224500000416
The inter-lead characteristics of 12-lead electrocardiosignals are respectively output after being input into the second branch unit
Figure FDA00036427224500000417
And features of
Figure FDA00036427224500000418
c-5.3) the third branch unit of the fourth residual structure is composed of a convolution layer and a Maxpool layer in sequence, and the characteristics are respectively
Figure FDA00036427224500000419
And features
Figure FDA00036427224500000420
Respectively outputting the signals to obtain the characteristics of 12-lead electrocardiosignals after being input into a third branch unit
Figure FDA00036427224500000421
And features of
Figure FDA00036427224500000422
c-5.4) characterization of leads
Figure FDA00036427224500000423
And features between leads
Figure FDA00036427224500000424
Inputting the spliced features into a Maxpool layer to output the obtained features
Figure FDA00036427224500000425
Inner features of leads
Figure FDA00036427224500000426
And features between leads
Figure FDA00036427224500000427
After feature splicing, inputting the feature into a Maxpool layer of the Maxpool layer and outputting the feature to obtain features
Figure FDA00036427224500000428
c-5.5) characterization of
Figure FDA00036427224500000429
And features of
Figure FDA00036427224500000430
Overlap to form new features
Figure FDA00036427224500000431
Will be characterized by
Figure FDA00036427224500000432
And features of
Figure FDA00036427224500000433
Performing superposition to form new features
Figure FDA00036427224500000434
c-6) characterization of
Figure FDA00036427224500000435
And features of
Figure FDA00036427224500000436
Inputting the data into a full-connection classifier, and outputting the data to obtain a label predicted value of a source domain
Figure FDA00036427224500000437
And label prediction value of target domain
Figure FDA00036427224500000438
4. The method for classifying multi-target domain electrocardiosignals based on depth domain self-adaptation according to claim 1, which is characterized in that: in the step c-2.1), the number of channels of the first convolution layer and the second convolution layer of the first branch unit is 32, the size of convolution kernels is 1 × 3, in the step c-2.2), the number of channels of the convolution layer of the second branch unit is 32, the size of convolution kernels is 1 × 3, in the DeformConv variable convolution layer, the number of channels is 32, the size of convolution kernels is 3 × 3, in the step c-2.3), the size of convolution kernels of the Maxpool maximum pooling layer is 1 × 4, and in the step c-2.4), the size of convolution kernels of the Maxpool maximum pooling layer is 1 × 4.
5. The method for classifying multi-target domain electrocardiosignals based on depth domain self-adaptation according to claim 1, which is characterized in that: in the step c-3.1), the number of channels of the first convolution layer and the second convolution layer of the first branch unit is 64, the size of convolution kernels is 1 × 3, the number of channels of the convolution layer of the second branch unit in the step c-3.2) is 64, the size of convolution kernels is 1 × 3, the number of channels of the DeformConv variable convolution layer is 64, the size of convolution kernels is 3 × 3, the size of convolution kernels of the Maxpool maximum pooling layer in the step c-3.3) is 1 × 4, and the size of convolution kernels of the Maxpool maximum pooling layer in the step c-3.4) is 1 × 4.
6. The method for classifying multi-target domain electrocardiosignals based on depth domain self-adaptation according to claim 1, which is characterized in that: the number of channels of the first convolution layer and the second convolution layer of the first branch unit in the step c-4.1) is 128, the size of convolution kernels is 1 x 3, the number of channels of the convolution layer of the second branch unit in the step c-4.2) is 128, the size of convolution kernels is 1 x 3, the number of channels of the DeformConv variable convolution layer is 128, the size of convolution kernels is 3 x 3, the size of convolution kernels of the Maxpool maximum pooling layer in the step c-4.3) is 1 x 4, and the size of convolution kernels of the Maxpool maximum pooling layer in the step c-4.4) is 1 x 4.
7. The method for classifying multi-target domain electrocardiosignals based on depth domain self-adaptation according to claim 1, which is characterized in that: the number of channels of the first convolution layer and the second convolution layer of the first branch unit in the step c-5.1) is 256, the sizes of convolution kernels are 1 × 3, the number of channels of the convolution layer of the second branch unit in the step c-5.2) is 256, the sizes of convolution kernels are 1 × 3, the number of channels of the DeformConv variable convolution layer is 256, the sizes of convolution kernels are 3 × 3, the size of convolution kernel of the Maxpool layer in the step c-5.3) is 1 × 4, and the size of convolution kernel of the Maxpool layer in the step c-5.4) is 1 × 4.
8. The method for classifying multi-target domain electrocardiosignals based on depth domain self-adaptation according to claim 1, wherein the step d) comprises the following steps:
d-1) by the formula
Figure FDA0003642722450000051
Calculating to obtain cross entropy classification loss
Figure FDA0003642722450000061
In the formula
Figure FDA0003642722450000062
N S Is the total number of samples in the source domain S,
Figure FDA0003642722450000063
is the number of samples with class j in the source domain S, j is 1,2 b In order to train the batch size value,
Figure FDA0003642722450000064
for the true tag value of the ith sample in the source domain S,
Figure FDA0003642722450000065
the label prediction value of the ith sample in the source domain S is obtained;
d-2) by the formula
Figure FDA0003642722450000066
Computationally derived class certainty loss
Figure FDA0003642722450000067
Where C is the number of all classes, ω 1 And omega 2 Are sample deterministic weighting coefficients, T is a transpose,
Figure FDA0003642722450000068
tag prediction values for target domains
Figure FDA0003642722450000069
Prediction probability matrix obtained by Softmax
Figure FDA00036427224500000610
The (c) th column of (a),
Figure FDA00036427224500000611
tag prediction values for target domains
Figure FDA00036427224500000612
Prediction by SoftmaxProbability matrix
Figure FDA00036427224500000613
Column j';
d-3) by the formula
Figure FDA00036427224500000614
Calculating to obtain class diversity loss
Figure FDA00036427224500000615
Figure FDA00036427224500000616
To predict a probability matrix
Figure FDA00036427224500000617
The rank of (d);
d-4) by the formula
Figure FDA00036427224500000618
Calculating a loss function
Figure FDA00036427224500000619
Where α is a coefficient and β is a coefficient.
CN202210526548.4A 2022-05-13 2022-05-13 Multi-target domain electrocardiosignal classification method based on depth field self-adaption Pending CN114841209A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210526548.4A CN114841209A (en) 2022-05-13 2022-05-13 Multi-target domain electrocardiosignal classification method based on depth field self-adaption

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210526548.4A CN114841209A (en) 2022-05-13 2022-05-13 Multi-target domain electrocardiosignal classification method based on depth field self-adaption

Publications (1)

Publication Number Publication Date
CN114841209A true CN114841209A (en) 2022-08-02

Family

ID=82569616

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210526548.4A Pending CN114841209A (en) 2022-05-13 2022-05-13 Multi-target domain electrocardiosignal classification method based on depth field self-adaption

Country Status (1)

Country Link
CN (1) CN114841209A (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116778969A (en) * 2023-06-25 2023-09-19 山东省人工智能研究院 Domain-adaptive heart sound classification method based on double-channel cross attention
CN117257322A (en) * 2023-09-21 2023-12-22 齐鲁工业大学(山东省科学院) Multi-label electrocardiosignal classification method based on dual-branch network

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116778969A (en) * 2023-06-25 2023-09-19 山东省人工智能研究院 Domain-adaptive heart sound classification method based on double-channel cross attention
CN116778969B (en) * 2023-06-25 2024-03-01 山东省人工智能研究院 Domain-adaptive heart sound classification method based on double-channel cross attention
CN117257322A (en) * 2023-09-21 2023-12-22 齐鲁工业大学(山东省科学院) Multi-label electrocardiosignal classification method based on dual-branch network
CN117257322B (en) * 2023-09-21 2024-04-19 齐鲁工业大学(山东省科学院) Multi-label electrocardiosignal classification method based on dual-branch network

Similar Documents

Publication Publication Date Title
CN114841209A (en) Multi-target domain electrocardiosignal classification method based on depth field self-adaption
CN108764317B (en) Residual convolutional neural network image classification method based on multipath feature weighting
CN109190464B (en) Mechanical fault intelligent diagnosis method based on transfer learning under variable working conditions
CN110728224A (en) Remote sensing image classification method based on attention mechanism depth Contourlet network
CN112561910A (en) Industrial surface defect detection method based on multi-scale feature fusion
CN109190511B (en) Hyperspectral classification method based on local and structural constraint low-rank representation
CN112465069B (en) Electroencephalogram emotion classification method based on multi-scale convolution kernel CNN
CN112381763A (en) Surface defect detection method
CN114612715A (en) Edge federal image classification method based on local differential privacy
CN114355240A (en) Power distribution network ground fault diagnosis method and device
CN111553186A (en) Electromagnetic signal identification method based on depth long-time and short-time memory network
CN114636975A (en) LPI radar signal identification method based on spectrogram fusion and attention mechanism
CN111010222B (en) Deep learning-based large-scale MIMO downlink user scheduling method
CN113076920B (en) Intelligent fault diagnosis method based on asymmetric domain confrontation self-adaptive model
CN111768792A (en) Audio steganalysis method based on convolutional neural network and domain confrontation learning
CN113109782B (en) Classification method directly applied to radar radiation source amplitude sequence
CN113052099B (en) SSVEP classification method based on convolutional neural network
CN114359167A (en) Insulator defect detection method based on lightweight YOLOv4 in complex scene
CN113221614A (en) Power grid insulator damage image identification method based on hybrid neural network
CN116312860B (en) Agricultural product soluble solid matter prediction method based on supervised transfer learning
CN113239949A (en) Data reconstruction method based on 1D packet convolutional neural network
CN117171544A (en) Motor vibration fault diagnosis method based on multichannel fusion convolutional neural network
CN115035408A (en) Unmanned aerial vehicle image tree species classification method based on transfer learning and attention mechanism
CN111191548B (en) Discharge signal identification method and identification system based on S transformation
CN115062690A (en) Bearing fault diagnosis method based on domain adaptive network

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