One-dimensional electrocardiogram data classification method based on residual network
A classification method and ECG data technology, applied in the computer field, can solve the problems that the accuracy rate of ECG data classification is difficult to further improve, the classification accuracy rate varies greatly, and the training takes a long time, so as to achieve short training time and avoid The effect of overfitting and good generalization
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
Method used
Image
Examples
Embodiment 1
[0029] The present invention provides a one-dimensional electrocardiographic data classification method based on residual network, such as figure 1 As shown, the method includes the following steps:
[0030] S1, obtain ECG data and preprocess it;
[0031] S11, acquisition of ECG data;
[0032]The ECG data obtained by the present invention may be the ECG data directly collected from the subject, or may be the ECG data in an existing database. ECG data. More specifically, the ECG data of the embodiments of the present invention are derived from four major world ECG databases, including the MIT-BIH ECG data-arrhythmia database, the sudden cardiac death dynamic ECG database, and the EU ST-T database. ECG database, MIT-BIH ECG data-ST segment database. In order to use the data of different leads for training and ensure the training effect, the data in the EU ST-T ECG database is divided into EU ST-T ECG data-MLII lead data set, EU ST-T ECG data set, and EU ST-T ECG data set. D...
Embodiment 2
[0050] The difference between this embodiment and Embodiment 1 is that the network model constructed in step S2 is the classic ResNet-18 network, specifically, the method disclosed by the applicant according to the document entitled "Deep Residual Learning for Image Recognition" That is, it is built with relevant parameters and optimized only with the goal of tuning hyperparameters. More specifically, as Image 6 As shown, it includes convolutional layer 1+BN layer+ReLU+global average pooling layer, residual module, global maximum pooling layer, and fully connected layer in turn. More specifically, as Figure 7 As shown, the ECG data input to the network model passes through the convolution layer 1+BN layer+ReLU+global average pooling layer, the number of channels increases and the number of features decreases. Specifically, the number of channels changes from 1 to 64, and the number of features becomes half; After passing through the residual module, the number of channels ...
Embodiment 3
[0054] The difference between this embodiment and Embodiment 1 is that the network model constructed in step S2 is constructed based on the network model and related parameters disclosed in the document named "Deep Residual Learning for Image Recognition" which has been published in the journal IEEE. , a convolutional neural network model constructed with the optimization goal of adjusting the number of convolutional layers and the number of neurons in each hidden layer. As shown in FIG. 8 , the network model of this embodiment sequentially includes three “convolutional layer 11+BN layer+ReLU+global maximum pooling layer”, convolutional layer 12, and fully connected layer which are connected in sequence. Specifically, as Figure 9 As shown in the figure, the ECG data is input into the network model, and after three "convolutional layer 11 + BN layer + ReLU + global maximum pooling layer" and convolutional layer 12, the number of features is reduced and the number of channels i...
PUM
Login to View More Abstract
Description
Claims
Application Information
Login to View More