Classification And Implementation Of Arrhythmia Based On Deep Learning

Electrocardiogram is the most commonly used method for diagnosing cardiovascular disease.Because traditional ECG classification technology relies too much on artificial feature extraction,ECG auto-analysis cannot be achieved very well.Finally,the clinician needs to make a final judgment on the result.In recent years,with the advancement of science and technology and the development of artificial intelligence,it has become possible to apply deep learning technology to the automatic classification of ECG arrhythmia.This article combines deep learning techniques with electrocardiogram diagnosis to study the automatic classification of arrhythmia.The main research work is divided into the following aspects:Firstly,the digital filtering method and wavelet decomposition method are designed to filter the database ECG data.The ECG signal doped with noise is not conducive to the analysis of arrhythmia disease.According to the useful information and the distribution of the main noise frequency range,the median filtering and wavelet are designed.The denoising method combines three main noises.The heart beat was divided into a filtered arrhythmia database(Massachusetts Institute of Technology-Beth Israel Hospital,MIT-BIH)to produce a data set.Then,the shallow convolutional neural network mini4 and Vgg convolutional neural network were designed to classify the five arrhythmia data.The parameters of the network model are adjusted by the stochastic gradient descent method and the backpropagation algorithm to generate the corresponding network model.The test set classification results are given in the form of confusion matrix.Finally,the Vgg convolutional neural network is improved,and the Vgg-connect network is designed to classify arrhythmia.The confusion matrix is evaluated by sensitivity and positive detection rate index.Finally,the Vgg convolutional neural network model Vgg-connect is improved to reduce network parameters.Reducing the amount of network computing while maintaining roughly the same overall accuracy as the Vgg network.And the ability to correctly detect disease increased by 0.35%(mean positive detection rate),the average sensitivity of Vgg-connect for five arrhythmia classification results was 96.74%,the average positive detection rate was 96.96%,and the overall accuracy rate reached 97.51%.The Vgg-connect neural network studied in this paper satisfies the need for automatic classification of arrhythmia.Finally,Vgg-connect neural network is selected as the network model to realize automatic classification of arrhythmia.