Research On ECG Automatic Classification Algorithm Based On Deep Learning Technology

Cardiovascular disease is one of the diseases leading to higher mortality,and Electrocardiogram(ECG)is an effective non-invasive diagnostic tool commonly used to screen and diagnose cardiovascular disease.However,due to the large amount of dynamic electrocardiogram data and the limited number of medical experts,the diagnosis of doctors is very arduous.The use of computer-assisted ECG analysis tools can greatly reduce the workload of doctors and improve the efficiency of screening and diagnosis of cardiovascular disease.This paper aims to establish an automatic classification model of ECG data based on deep learning technology combined with the time series characteristics of ECG data.The main research contents of this paper are as follows:(1)For the problem of noise in ECG,which leads to difficulty in feature extraction,this paper uses wavelet transform algorithm to reduce the noise of ECG signal.According to the characteristics of ECG data,the Dauchechies6(db6)wavelet function is used to decompose the ECG signal into eight layers.Based on this,the reasonable threshold is set according to the noise figure for noise reduction.Finally,the threshold wavelets are processed.Structure,get the ECG signal after noise reduction.The experimental results show that the wavelet transform has a good noise reduction effect.(2)Aiming at the problem that traditional machine learning algorithms rely heavily on manual extraction of features,this paper optimizes and improves the convolutional neural network model in deep learning technology,and designs an efficient convolutional neural network E-CNN(Efficient convolutional neural network)for single Lead ECG is automatically classified.E-CNN can extract multi-level features of ECG data from the same input,and can efficiently obtain internal structural feature representation of ECG data.The experimental results show that E-CNN has good classification performance in ECG classification.For the multi-lead ECG of two-dimensional structure,this paper proposes a multi-channel convolutional neural network(MC-CNN).The MC-CNN modelautomatically imports each lead in the multi-lead ECG data into different channels.The multi-channel design of the MC-CNN model not only ensures the independence of the data between the leads,but also enables each The lead finds a filter that suits you and extracts high quality ECG features.The experimental results show that the MC-CNN model has great advantages in the automatic classification of multi-lead ECG.Based on the characteristics of ECG data,two ECG automatic classification models E-CNN and MC-CNN based on deep learning technology were designed and validated on MIT-BIH arrhythmia dataset and PTB myocardial infarction dataset..Experiments show that the proposed model not only solves the problem of manual extraction of features,but also extracts high-quality ECG features,and has obtained good ECG automatic classification results.