Research On Arrhythmia Classification Technology Based On Data-driven

Heart disease is mostly a chronic.Chronic cardiovascular diseases have the characteristics of high morbidity,high disability and high mortality.Among them,arrhythmia is a more common type of cardiovascular disease.Most cardiovascular diseases are often accompanied by arrhythmias.In clinical,electrocardiogram(ECG)can be used as the main diagnostic tool for cardiac activity,and it is often used to detect arrhythmia.ECG signal has the characteristics of concealment and burst,and the signal amplitude is small.Traditional visual inspection will miss important information of ECG,which will affect the accuracy of disease diagnosis.In order to improve the efficiency of medical diagnosis,shorten the time of diagnosis and improve the rate of disease recognition,computer-aided diagnosis is introduced into ECG analysis.The basis of ECG data classification and recognition technology is to extract features effectively.There are individual differences in ECG waveforms,inaccurate features and missing useful features,which can not effectively extract the hidden features behind the massive ECG signals.Deep learning method integrates feature extraction and classification,avoids partial deviation in the process of feature extraction,and provides a new idea for automatic diagnosis of cardiovascular diseases.How to classify and recognize ECG time series data effectively is a common problem in the auxiliary diagnosis of cardiovascular diseases.Based on the characteristics of ECG data sequence,this paper aims to assist the diagnosis of cardiovascular diseases,and studies the spatial and temporal characteristics of ECG data,and builds an automatic classification model of ECG data by using the deep learning method.The main research content of the paper includes:(1)According to the characteristics of time series correlation of ECG data,a DRCNN-Bi GRU(Deep Residual CNN-Bi GRU)hybrid model is constructed to study the deep-seated nature of the target data.Combined with the translation invariance of the convolutional network and the characteristics of the recursive network context information processing mechanism,the timing and spatial characteristics of ECG data are automatically extracted.Firstly,the median filter and band stop filter are used to preprocess ECG data;secondly,the DRCNN-Bi GRU hybrid model is used to independently learn the spatial and temporal order characteristics of ECG,and classify the arrhythmia types of different sampling points;finally,based on the MIT-BIH arrhythmia database,DRCNN-Bi GRU verifies six kinds of arrhythmias when the data segment length is 400 sampling points The accuracy of the model is 99.59%.(2)Aiming at the difference of feature importance in ECG time series data,a hybrid model of SE-DRCNN-Bi GRU is constructed by introducing se block into deep neural network.Based on the characteristics of local features captured by deep CNN and timing information captured by Bi GRU,combined with the importance of each feature channel in the se block learning network with attention thought,the SEDRCNN-Bi GRU hybrid model is constructed for ECG classification and recognition,to solve the problem of feature loss caused by the different importance of different channels in the convolution pooling process.The model was validated by using MITBIH-AF(MIT-BIH Atrial Fibrillation,MIT-BIH-AF)atrial fibrillation ECG database and CCDD(Chinese cardiovascular,disease database,CCDD)clinical experiment database.The results show that the accuracy of classification and recognition of the model is 98.78% and 90.97%.(3)Design and implement remote ECG monitoring system.For the target users of remote ECG monitoring,the overall framework and function modules are designed according to the needs analysis of the system,and the corresponding functional modules are implemented.Based on the above-mentioned auxiliary diagnosis model,remote ECG-assisted diagnosis is implemented to verify the feasibility and reliability of the algorithm.Design and implement remote ECG monitoring system.For the target user of remote ECG monitoring,the overall framework and functional module design are analyzed according to the needs of the system to implement the corresponding system functional module and complete the test.Based on the above-mentioned auxiliary diagnosis model,remote ECG-assisted diagnosis is implemented to verify the feasibility and reliability of the algorithm.