Research On Arrhythmia Classification Based On Convolutional Neural Network

As our country gradually enters an aging society,the number of patients with cardiovascular diseases continues to increase,making the demand of the cardiac monitoring system continue to increase.How to accurately detect and classify arrhythmia with computer aided technology is one of the hot spots in the diagnosis of cardiovascular diseases.When recognizing arrhythmia,the cardiologist traditionally observes and analyzes the patients’ s electrocardiogram which tends to lead to missed detection and objective conclusion.Therefore,this thesis considers using machine learning and deep learning classification algorithms to conduct the following research on arrhythmia:Firstly,in view of the classification of arrhythmias in patients with a single heartbeat,two classification models based on traditional machine learning method support vector machine and deep learning method based on convolutional neural network are proposed to classify four categories of electrocardiogram.The data preprocessing methods of cutting electrocardiogram signals into heart beats were input into the classification model for experimental comparative analysis.The experimental results show that the convolutional neural network model for arrhythmia classification can automatically extract the features of electrocardiogram data,which reduces the steps of classification and has a good classification effect.Secondly,aiming at the classification of arrhythmias among patients with continuous electrocardiogram sequence,an attention residual network model was proposed to classify arrhythmias into five categories.The method of attention mechanism is introduced into the residual network model to make it easy to capture the important features of electrocardiogram signals,and then the classification model is studied for ablation.The overfitting problem of the proposed network model was optimized by data enhancement of the original electrocardiogram data.Experimental results show that the proposed algorithm is feasible and enhances the robustness of the network model.Finally,for the detection of atrial fibrillation in the CINC2017(Computing IN Cardiology Challenge 2017,CINC2017)data set,a nine-layer convolutional neural network model was constructed in the experiment and a new data preprocessing method was proposed to identify atrial fibrillation and classify normal rhythm,other rhythm and noisy electrocardiogram signals.The analysis of experimental results shows that the proposed network model can improve the classification accuracy of atrial fibrillation.