CN114886437A - Ventricular premature beat identification method based on improved ShuffleNetV2 - Google Patents
Ventricular premature beat identification method based on improved ShuffleNetV2 Download PDFInfo
- Publication number
- CN114886437A CN114886437A CN202210508517.6A CN202210508517A CN114886437A CN 114886437 A CN114886437 A CN 114886437A CN 202210508517 A CN202210508517 A CN 202210508517A CN 114886437 A CN114886437 A CN 114886437A
- Authority
- CN
- China
- Prior art keywords
- layer
- output
- input
- unit block
- premature beat
- 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
- 208000000418 Premature Cardiac Complexes Diseases 0.000 title claims abstract description 49
- 230000002861 ventricular Effects 0.000 title claims abstract description 43
- 238000000034 method Methods 0.000 title claims abstract description 12
- 230000007246 mechanism Effects 0.000 claims abstract description 34
- 238000012549 training Methods 0.000 claims abstract description 14
- 238000011176 pooling Methods 0.000 claims description 21
- 230000004913 activation Effects 0.000 claims description 15
- 238000007781 pre-processing Methods 0.000 claims description 9
- 238000005096 rolling process Methods 0.000 claims description 9
- 230000002028 premature Effects 0.000 claims description 3
- 230000003993 interaction Effects 0.000 abstract description 2
- 230000006870 function Effects 0.000 description 10
- 238000010586 diagram Methods 0.000 description 4
- 230000035945 sensitivity Effects 0.000 description 4
- 210000004027 cell Anatomy 0.000 description 3
- 238000013136 deep learning model Methods 0.000 description 3
- 238000010606 normalization Methods 0.000 description 3
- 208000024172 Cardiovascular disease Diseases 0.000 description 2
- 206010003119 arrhythmia Diseases 0.000 description 2
- 230000006793 arrhythmia Effects 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 238000012935 Averaging Methods 0.000 description 1
- 206010015856 Extrasystoles Diseases 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000015556 catabolic process Effects 0.000 description 1
- 238000006731 degradation reaction Methods 0.000 description 1
- 201000010099 disease Diseases 0.000 description 1
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000001629 suppression Effects 0.000 description 1
Images
Classifications
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
- A61B5/316—Modalities, i.e. specific diagnostic methods
- A61B5/318—Heart-related electrical modalities, e.g. electrocardiography [ECG]
- A61B5/346—Analysis of electrocardiograms
- A61B5/349—Detecting specific parameters of the electrocardiograph cycle
- A61B5/364—Detecting abnormal ECG interval, e.g. extrasystoles, ectopic heartbeats
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/68—Arrangements of detecting, measuring or recording means, e.g. sensors, in relation to patient
- A61B5/6801—Arrangements of detecting, measuring or recording means, e.g. sensors, in relation to patient specially adapted to be attached to or worn on the body surface
- A61B5/6802—Sensor mounted on worn items
- A61B5/681—Wristwatch-type devices
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/048—Activation functions
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Health & Medical Sciences (AREA)
- Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Physics & Mathematics (AREA)
- Molecular Biology (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Theoretical Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Cardiology (AREA)
- General Physics & Mathematics (AREA)
- Medical Informatics (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Data Mining & Analysis (AREA)
- Pathology (AREA)
- Heart & Thoracic Surgery (AREA)
- Computing Systems (AREA)
- Surgery (AREA)
- Animal Behavior & Ethology (AREA)
- Public Health (AREA)
- Veterinary Medicine (AREA)
- Computational Linguistics (AREA)
- Physiology (AREA)
- Psychiatry (AREA)
- Signal Processing (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)
Abstract
The invention discloses a ventricular premature beat identification method based on an improved ShuffleNet V2, and provides a ventricular premature beat identification model based on an improved ShuffleNet V2 model. Training the constructed ventricular premature beat recognition model by using the training samples, and then sending the prediction samples into the trained ventricular premature beat recognition model to realize the classification of the electrocardiogram to be classified. The ECA attention mechanism module learns effective channel attention by avoiding reducing channel dimensions while capturing interaction information between different channels in a very lightweight manner. Through experimental comparison, the ShuffeNet V2 model with the ECA attention mechanism module obtains indexes higher than those of the ShuffeNet V2 model without the ECA attention mechanism module. The improved ShuffleNet V2 has the advantages of high precision, high speed and the like, and is easy to deploy in wearable devices such as a wristwatch, a bracelet and the like.
Description
Technical Field
The invention relates to the technical field of Ventricular Premature beat identification, in particular to a Ventricular Premature beat (PVC) identification method based on improved ShuffleNet V2.
Background
Arrhythmia is a common cardiovascular disease in clinic, and premature beat is the most common arrhythmia. By introducing the computer-assisted premature beat analysis technology, doctors can perform effective medical intervention as early as possible, play an important role in preventing and treating diseases, and can solve the problem that basic medical services cannot be enjoyed due to the fact that medical resources in remote mountainous areas are scarce to a certain extent.
For the deep learning model, besides the accuracy, the computation complexity is also an important index to be considered by the deep learning model. If the model is too complex, the running speed of the model can be slow, and the model is difficult to deploy into a mobile end, such as a wearable device like a wristwatch and a bracelet.
Disclosure of Invention
The invention aims to solve the problem that the existing deep learning model has deep layers and complex network, so that the model deployment to a mobile platform is influenced, and provides an improved ShuffleNet V2-based ventricular premature beat identification method.
In order to solve the problems, the invention is realized by the following technical scheme:
the ventricular premature beat identification method based on the improved ShuffleNetV2 comprises the following steps:
the input of the first convolution layer is used as the input of the ventricular premature beat recognition model; the output of the first rolling layer is connected with the input of the maximum pooling layer, the output of the maximum pooling layer is connected with the input of the first unit block, the output of the first unit block is connected with the input of the first second unit block, the output of the first second unit block is connected with the input of the second first unit block, the output of the second unit block is connected with the input of the third second unit block, the output of the third second unit block is connected with the input of the fourth second unit block, the output of the fourth second unit block is connected with the input of the third first unit block, the input of the third first unit block is connected with the input of the fifth second unit block, the output of the fifth second unit block is connected with the input of the second rolling layer, the output of the second rolling layer is connected with the input of the global average pooling layer, and the output of the global average pooling layer is connected with the input of the output layer; the output of the output layer is used as the output of the ventricular premature beat identification model;
and 3, preprocessing the electrocardiogram to be classified to be used as a prediction sample, and sending the prediction sample into the ventricular premature beat recognition model trained in the step 2 to realize the classification of the electrocardiogram to be classified.
In the scheme, the first unit block consists of a channel splitting layer, 2 convolution layers, 1 depth separable convolution layer, 1 ECA attention mechanism layer, 1 splicing layer and 1 channel disordering layer; the input of the channel splitting layer is used as the input of a first unit block; the output of the channel splitting layer is divided into two paths: one path is directly connected with one input of the first splicing layer; the other path of the output of the third convolution layer is connected with the input of the first depth separable convolution layer after passing through the batch standardization BN and the Relu activation function, the output of the first depth separable convolution layer is connected with the input of the first ECA attention mechanism layer after passing through the batch standardization BN, the output of the first ECA attention mechanism layer is connected with the input of the fourth convolution layer, and the output of the fourth convolution layer is connected with the other input of the first splicing layer after passing through the batch standardization BN and the Relu activation function; the output of the first splicing layer is connected with the input of the first channel scrambling layer, and the output of the first channel scrambling layer is used as the output of the first unit block.
In the scheme, the second unit block consists of 3 convolution layers, 2 depth separable convolution layers, 2 ECA attention mechanism layers, 1 splicing layer and 1 channel disordering layer; the inputs of the second depth separable convolutional layer and the sixth convolutional layer are used as the input of the unit block II together; the output of the second depth separable convolutional layer is connected with the input of the second ECA attention mechanism layer after passing through the batch standardized BN, the output of the second ECA attention mechanism layer is connected with the input of the fifth convolutional layer, and the output of the fifth convolutional layer is connected with one input of the second splicing layer after passing through the batch standardized BN and the Relu activation function; the output of the sixth convolutional layer is connected with the input of the third depth separable convolutional layer after passing through the batch standardization BN and the Relu activation function, the output of the third depth separable convolutional layer is connected with the third ECA attention mechanism layer after passing through the batch standardization BN, the output of the third ECA attention mechanism layer is connected with the input of the seventh convolutional layer, and the output of the seventh convolutional layer is connected with the other input of the second splicing layer after passing through the batch standardization BN and the Relu activation function; and the output of the second splicing layer is connected with the input of a second channel scrambling layer, and the output of the second channel scrambling layer is used as the output of the second cell block.
In the above scheme, the step size of the convolution kernel of the depth separable convolution layer is 1.
Compared with the prior art, the invention has the following characteristics:
1. the improved ShuffleNet V2 has the advantages of high precision, high speed and the like, and is easy to deploy in wearable devices such as a wristwatch, a bracelet and the like.
2. The ECA attention mechanism module learns effective channel attention by avoiding degradation of channel dimensions while capturing interaction information between different channels in a very lightweight manner. Through experimental comparison, the ShuffeNet V2 model with the ECA attention mechanism module obtains indexes higher than those of the ShuffeNet V2 model without the ECA attention mechanism module.
Drawings
Fig. 1 is a schematic structural diagram of a ventricular premature beat identification model based on the improved ShuffleNet V2.
FIG. 2 is a schematic diagram of Block1 in FIG. 1.
FIG. 3 is a schematic diagram of Block2 in FIG. 1.
FIG. 4 is a schematic diagram of the structure of the ECA attention suppression layer of FIG. 2.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is further described in detail with reference to specific examples.
A ventricular premature beat identification method based on improved shefflenetv 2, comprising the following steps:
Considering that a mobile terminal device usually needs a small model which is accurate and fast, the invention carries out related research on a lightweight network model ShuffleNet V2, improves the ShuffleNet V2 based on ventricular premature beat identification characteristics, and provides a ventricular premature beat identification model which is composed of 2 convolutional layers (Conv), 1 maximum pooling layer (MaxPool), 3 basic units (Stage), 1 global averaging pooling layer and an output layer as shown in FIG. 1. Each basic unit is composed of a unit Block of one number (Block1) and a unit Block of two numbers (Block 2). The first basic unit consists of 1 first unit block and 1 second unit block; the second basic unit consists of 1 first unit block and 3 second unit blocks; the third basic unit is composed of 1 unit block of the first number and 1 unit block of the second number. The input of the first convolutional layer serves as the input of the ventricular premature beat recognition model. The output of the first rolling layer is connected with the input of the maximum pooling layer, the output of the maximum pooling layer is connected with the input of the first unit block, the output of the first unit block is connected with the input of the first second unit block, the output of the first second unit block is connected with the input of the second first unit block, the output of the second unit block is connected with the input of the third second unit block, the output of the third second unit block is connected with the input of the fourth second unit block, the output of the fourth second unit block is connected with the input of the third first unit block, the input of the third first unit block is connected with the input of the fifth second unit block, the output of the fifth second unit block is connected with the input of the second rolling layer, the output of the second rolling layer is connected with the input of the global average pooling layer, and the output of the global average pooling layer is connected with the input of the output layer. The output of the output layer is used as the output of the ventricular premature beat identification model.
Referring to fig. 2, the first cell block is composed of a Channel Split layer (Channel Split), 2 convolutional layers (Conv), 1 depth separable convolutional layer (DWConv), 1 ECA attention mechanism layer, 1 splice layer (Concat), and 1 Channel shuffle layer (Channel shuffle). The input of the channel split layer is used as the input of the unit block number one. The output of the channel splitting layer is divided into two paths: one path is directly connected with one input of the first splicing layer; the other path is connected with the input of a third convolution layer, the output of the third convolution layer is connected with the input of a first depth separable convolution layer after passing through a batch standardization BN and a Relu activation function, the output of the first depth separable convolution layer is connected with the input of a first ECA attention mechanism layer after passing through the batch standardization BN, the output of the first ECA attention mechanism layer is connected with the input of a fourth convolution layer, and the output of the fourth convolution layer is connected with the other input of the first splicing layer after passing through the batch standardization BN and the Relu activation function. The output of the first splicing layer is connected with the input of the first channel scrambling layer, and the output of the first channel scrambling layer is used as the output of the first unit block.
Referring to fig. 3, the unit block No. two is composed of 3 convolutional layers (Conv), 2 depth separable convolutional layers (DWConv), 2 ECA attention mechanism layers, 1 splice layer (Concat), and 1 channel shuffle layer (channel shuffle). The inputs of the second depth-separable convolutional layer and the sixth convolutional layer are collectively the input of the unit block number two. The output of the second depth separable convolutional layer is connected to the input of the second ECA attention mechanism layer after passing through the batch normalized BN, the output of the second ECA attention mechanism layer is connected to the input of the fifth convolutional layer, and the output of the fifth convolutional layer is connected to one input of the second stitching layer after passing through the batch normalized BN and the Relu activation function. The output of the sixth convolutional layer is connected with the input of the third depth separable convolutional layer after passing through the batch normalization BN and the Relu activation function, the output of the third depth separable convolutional layer is connected with the third ECA attention mechanism layer after passing through the batch normalization BN, the output of the third ECA attention mechanism layer is connected with the input of the seventh convolutional layer, and the output of the seventh convolutional layer is connected with the other input of the second splicing layer after passing through the batch normalization BN and the Relu activation function. And the output of the second splicing layer is connected with the input of a second channel scrambling layer, and the output of the second channel scrambling layer is used as the output of the second cell block.
In the first unit block and the second unit block, all the convolution layers are 1 × 1 convolution layers, and all the depths can be separatedThe convolutional layers are all 1 x 3 depth separable convolutional layers. In the original ShuffleNetV2 model, depth separable convolution layers (DWConv) of a unit block one and a unit block two are downsampled, that is, the moving step Stride of the convolution kernel is 2. In the present invention, however, since DWConv requires that the convolution kernel is shifted by the same step size in both the horizontal and vertical directions, i.e., Stride _ x is equal to Stride _ y, the height of the ECG number one-dimensional signal data is 1; therefore, for the ventricular premature beat recognition model, the moving step size Stride of the convolution kernels of the first unit block and the second unit block is changed into 1, meanwhile, the 1-1 convolution layers of the first unit block and the second unit block need to pass through BN and Relu, and the 1-3 depth separable convolution layers only need to pass through BN. The purpose of the channel scrambling operation is to ensure that the information of the two branches is interacted. The ECA attention mechanism layer is shown in fig. 4, an input vector x of the ECA attention mechanism layer is processed by the ECA attention mechanism to obtain a weight value G of each neuron in the output and input vector x, and the weight value e is used t Multiplied by the input vector x to obtain the output y of the ECA attention mechanism layer.
And 2, preprocessing the electrocardiogram marked with the classification to be used as a training sample, and training the ventricular premature beat recognition model constructed in the step 1 by using the training sample to obtain the trained ventricular premature beat recognition model.
The preprocessing process of the ECG mainly utilizes a filter to remove noises such as baseline drift, power frequency interference and the like in the ECG. The categories marked in the electrocardiogram are either premature or non-premature.
And 3, preprocessing the electrocardiogram to be classified to be used as a prediction sample, and sending the prediction sample into the ventricular premature beat recognition model trained in the step 2 to realize the classification of the electrocardiogram to be classified.
The way of preprocessing the electrocardiogram to be classified is the same as the way of preprocessing the electrocardiogram with labeled classification. The output of the prediction sample obtained after passing through the model is a premature beat or non-premature beat record.
The performance of the invention is illustrated below by means of a specific example:
the parameters of the layers of the ventricular premature beat identification model of the present embodiment are shown in table 1, where KSize in table 1 represents the convolution kernel size of the convolutional layer (Conv1, Conv5) and the pooling kernel size of the maximum pooling layer (MaxPool), and Stride is the convolution kernel or the pooling kernel moving step. Each Stage is composed of blocks 1 and blocks 2, the specific number of blocks corresponds to Repeat columns, namely Stage2 passes through 1 Block1 and then 1 Block 2; stage3 passes through a Block1 and then 3 blocks 2; stage4 goes through 1 Block1 first and 1 Block2 second. Output channels represent the number of Output profiles per convolutional layer or the number of Output profiles used per Stage is Output channels/2.
TABLE 1 ShuffleNet V2 model in PVC identification
Taking a Chinese Cardiovascular Disease Database (CCDD) as a base, and carrying out denoising processing on the ECG records in the Database by 0.5-40 Hz band-pass filtering.
Since the premature beat of the present invention is classified as a two-classification problem (premature beat or non-premature beat), sensitivity (Se), specificity (Sp) and accuracy (Acc) can be used to measure the quality of classification. The confusion matrix for the second category is shown in table 1 below:
TABLE 1 confusion matrix
The definition of each index is as follows:
sensitivity (Se):
Se＝TP/(TP+FN)
specificity (Sp):
Sp＝TN/(TN+FP)
accuracy (Acc):
Acc＝(TP+TN)/(TP+TN+FP+FN)
35840 pre-processed ECG recordings were used as training samples, with 3112 ventricular premature beats (PVC); 141046 records (including 2148 PVC records) were used as test specimens. After training the improved ShuffleNet V2 model with the training samples, the results obtained for the model on 141046 recorded test samples are shown in Table 1 below.
TABLE 1PVC identification based on the improved ShuffLeNet V2 model
Wherein NPVC represents non-ventricular premature beat record, PVC is ventricular premature beat record, Se is sensitivity, Sp is specificity, and Acc is accuracy.
As can be seen from Table 1, the sensitivity, specificity and accuracy obtained by the classification of the improved ShuffLeNet V2 model all reach more than 90%.
It should be noted that, although the above-mentioned embodiments of the present invention are illustrative, the present invention is not limited thereto, and thus the present invention is not limited to the above-mentioned embodiments. Other embodiments, which can be made by those skilled in the art in light of the teachings of the present invention, are considered to be within the scope of the present invention without departing from its principles.
Claims (4)
1. The ventricular premature beat identification method based on the improved ShuffleNetV2 is characterized by comprising the following steps of:
step 1, constructing a ventricular premature beat identification model; the ventricular premature beat identification model consists of 2 convolutional layers, 1 maximum pooling layer, 3 basic units, 1 global average pooling layer and an output layer; the first basic unit consists of 1 first unit block and 1 second unit block; the second basic unit consists of 1 first unit block and 3 second unit blocks; the third basic unit consists of 1 first unit block and 1 second unit block;
the input of the first convolution layer is used as the input of the ventricular premature beat recognition model; the output of the first rolling layer is connected with the input of the maximum pooling layer, the output of the maximum pooling layer is connected with the input of the first unit block, the output of the first unit block is connected with the input of the first second unit block, the output of the first second unit block is connected with the input of the second first unit block, the output of the second unit block is connected with the input of the third second unit block, the output of the third second unit block is connected with the input of the fourth second unit block, the output of the fourth second unit block is connected with the input of the third first unit block, the input of the third first unit block is connected with the input of the fifth second unit block, the output of the fifth second unit block is connected with the input of the second rolling layer, the output of the second rolling layer is connected with the input of the global average pooling layer, and the output of the global average pooling layer is connected with the input of the output layer; the output of the output layer is used as the output of the ventricular premature beat identification model;
step 2, preprocessing the electrocardiogram marked with the classification to be used as a training sample, and training the ventricular premature beat recognition model constructed in the step 1 by using the training sample to obtain a trained ventricular premature beat recognition model;
and 3, preprocessing the electrocardiogram to be classified to be used as a prediction sample, and sending the prediction sample into the ventricular premature beat recognition model trained in the step 2 to realize the classification of the electrocardiogram to be classified.
2. The improved ShuffleNetV 2-based ventricular premature beat identification method according to claim 1, wherein the unit block number one consists of a channel splitting layer, 2 convolutional layers, 1 depth separable convolutional layer, 1 ECA attention mechanism layer, 1 splicing layer, and 1 channel scrambling layer;
the input of the channel splitting layer is used as the input of a first unit block; the output of the channel splitting layer is divided into two paths: one path is directly connected with one input of the first splicing layer; the other path of the output of the third convolution layer is connected with the input of the first depth separable convolution layer after passing through the batch standardization BN and the Relu activation function, the output of the first depth separable convolution layer is connected with the input of the first ECA attention mechanism layer after passing through the batch standardization BN, the output of the first ECA attention mechanism layer is connected with the input of the fourth convolution layer, and the output of the fourth convolution layer is connected with the other input of the first splicing layer after passing through the batch standardization BN and the Relu activation function; the output of the first splicing layer is connected with the input of the first channel scrambling layer, and the output of the first channel scrambling layer is used as the output of the first unit block.
3. The improved ShuffLeNet V2-based ventricular premature identification method according to claim 1, wherein the second unit block is composed of 3 convolutional layers, 2 depth separable convolutional layers, 2 ECA attention mechanism layers, 1 splice layer and 1 channel scrambling layer;
the input of the second depth separable convolutional layer and the input of the sixth convolutional layer are used as the input of the second unit block together; the output of the second depth separable convolutional layer is connected with the input of a second ECA attention mechanism layer after passing through the batch standardized BN, the output of the second ECA attention mechanism layer is connected with the input of a fifth convolutional layer, and the output of the fifth convolutional layer is connected with one input of a second splicing layer after passing through the batch standardized BN and the Relu activation function; the output of the sixth convolutional layer is connected with the input of the third depth separable convolutional layer after passing through the batch standardization BN and the Relu activation function, the output of the third depth separable convolutional layer is connected with the third ECA attention mechanism layer after passing through the batch standardization BN, the output of the third ECA attention mechanism layer is connected with the input of the seventh convolutional layer, and the output of the seventh convolutional layer is connected with the other input of the second splicing layer after passing through the batch standardization BN and the Relu activation function; and the output of the second splicing layer is connected with the input of a second channel scrambling layer, and the output of the second channel scrambling layer is used as the output of the second cell block.
4. The improved ShuffleNet V2-based ventricular premature beat identification method according to claim 2 or 3, wherein the step size of the movement of the convolution kernel of the depth separable convolution layer is 1.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210508517.6A CN114886437A (en) | 2022-05-10 | 2022-05-10 | Ventricular premature beat identification method based on improved ShuffleNetV2 |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210508517.6A CN114886437A (en) | 2022-05-10 | 2022-05-10 | Ventricular premature beat identification method based on improved ShuffleNetV2 |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114886437A true CN114886437A (en) | 2022-08-12 |
Family
ID=82720857
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210508517.6A Pending CN114886437A (en) | 2022-05-10 | 2022-05-10 | Ventricular premature beat identification method based on improved ShuffleNetV2 |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114886437A (en) |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106725420A (en) * | 2015-11-18 | 2017-05-31 | 中国科学院苏州纳米技术与纳米仿生研究所 | VPB recognition methods and VPB identifying system |
| CN109730652A (en) * | 2019-03-08 | 2019-05-10 | 江苏霆升科技有限公司 | Ventricular premature beat positioning identifying method and device |
| CN111772619A (en) * | 2020-06-19 | 2020-10-16 | 厦门纳龙科技有限公司 | Electrocardiogram heart beat identification method, terminal device and storage medium |
| CN112842349A (en) * | 2021-02-20 | 2021-05-28 | 无锡市中健科仪有限公司 | Intelligent analysis method for supraventricular premature beat based on attention transfer mechanism |
| CN112932433A (en) * | 2021-02-05 | 2021-06-11 | 青岛大学 | Arrhythmia automatic classification method of electrocardiogram |
| CN113712571A (en) * | 2021-06-18 | 2021-11-30 | 陕西师范大学 | Abnormal electroencephalogram signal detection method based on Rinyi phase transfer entropy and lightweight convolutional neural network |
| WO2022016638A1 (en) * | 2020-07-22 | 2022-01-27 | 上海数创医疗科技有限公司 | Training method and model for convolutional neural network based on sample distribution improvement |
| CN114419415A (en) * | 2022-01-11 | 2022-04-29 | 南通大学 | Improved digital identification method |
-
2022
- 2022-05-10 CN CN202210508517.6A patent/CN114886437A/en active Pending
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106725420A (en) * | 2015-11-18 | 2017-05-31 | 中国科学院苏州纳米技术与纳米仿生研究所 | VPB recognition methods and VPB identifying system |
| CN109730652A (en) * | 2019-03-08 | 2019-05-10 | 江苏霆升科技有限公司 | Ventricular premature beat positioning identifying method and device |
| CN111772619A (en) * | 2020-06-19 | 2020-10-16 | 厦门纳龙科技有限公司 | Electrocardiogram heart beat identification method, terminal device and storage medium |
| WO2022016638A1 (en) * | 2020-07-22 | 2022-01-27 | 上海数创医疗科技有限公司 | Training method and model for convolutional neural network based on sample distribution improvement |
| CN112932433A (en) * | 2021-02-05 | 2021-06-11 | 青岛大学 | Arrhythmia automatic classification method of electrocardiogram |
| CN112842349A (en) * | 2021-02-20 | 2021-05-28 | 无锡市中健科仪有限公司 | Intelligent analysis method for supraventricular premature beat based on attention transfer mechanism |
| CN113712571A (en) * | 2021-06-18 | 2021-11-30 | 陕西师范大学 | Abnormal electroencephalogram signal detection method based on Rinyi phase transfer entropy and lightweight convolutional neural network |
| CN114419415A (en) * | 2022-01-11 | 2022-04-29 | 南通大学 | Improved digital identification method |
Non-Patent Citations (2)
| Title |
|---|
| 周飞燕: "基于集成学习的室性早博识别方法", 《电子学报》, vol. 45, no. 2, 28 February 2017 (2017-02-28), pages 501 - 507 * |
| 王鹏: "基于深度学习的船舶驾驶员疲劳检测算法", 《交通信息与安全》, vol. 40, no. 1, 28 February 2022 (2022-02-28), pages 63 - 71 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110991236B (en) | Image classification method and related device | |
| CN110353673B (en) | Electroencephalogram channel selection method based on standard mutual information | |
| CN110991406A (en) | RSVP electroencephalogram characteristic-based small target detection method and system | |
| CN111599376B (en) | Sound event detection method based on cavity convolution cyclic neural network | |
| CN107909088A (en) | Obtain method, apparatus, equipment and the computer-readable storage medium of training sample | |
| Alqudah et al. | Lightweight Deep Learning for Malaria Parasite Detection Using Cell-Image of Blood Smear Images. | |
| CN114093501B (en) | Intelligent auxiliary analysis method for child movement epilepsy based on synchronous video and electroencephalogram | |
| CN111400536A (en) | Low-cost tomato leaf disease identification method based on lightweight deep neural network | |
| Pandya et al. | Unveiling the Power of Collective Intelligence: A Voting-based Approach for Dementia Classification | |
| CN116612335B (en) | Few-sample fine-granularity image classification method based on contrast learning | |
| CN115919330A (en) | EEG Emotional State Classification Method Based on Multi-level SE Attention and Graph Convolution | |
| CN114595725B (en) | Electroencephalogram signal classification method based on addition network and supervised contrast learning | |
| Razin et al. | Malaria parasite detection and classification using CNN and YOLOv5 architectures | |
| CN111833906A (en) | Sound scene classification method based on multi-path acoustic characteristic data enhancement | |
| CN114010205A (en) | Auxiliary analysis method for 3D (three-dimensional) attention residual error deep network children epilepsy syndrome | |
| CN117982102A (en) | Epileptic detection method and system based on cross-link nested subnetwork and mixed dependency algorithm | |
| CN117909822A (en) | Automatic epileptic detection method based on multi-scale attention | |
| CN114886437A (en) | Ventricular premature beat identification method based on improved ShuffleNetV2 | |
| Harrane et al. | Classification of ECG heartbeats using deep neural networks | |
| Yi et al. | Retinal artery/vein classification by multi-channel multi-scale fusion network | |
| CN115985331B (en) | Audio automatic analysis method for field observation | |
| CN114841216B (en) | Electroencephalogram signal classification method based on model uncertainty learning | |
| CN115500807B (en) | Arrhythmia classification detection method and system based on small convolutional neural network | |
| KR102290243B1 (en) | Classification method for Electroencephalography motor imagery | |
| CN115187629A (en) | Method for fusing target tracking features by using graph attention 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 |