Research On Atrial Fibrillation And First Degree Atrioventricular Block Identification Model Based On Machine Learning

With the increasing aging of the population and the prevalence of unhealthy lifestyle,the prevalence of various cardiovascular diseases in China is increasing day by day.There are about 330 million patients with cardiovascular diseases,and the mortality rate is still high.Atrial fibrillation(AF)and first degree atrioventricular block(i-avb)are two kinds of common arrhythmia diseases,which are closely related to many cardiovascular and cerebrovascular diseases.Therefore,the prevention,diagnosis and treatment of AF and i-avb are very important.With the development of Wearable ECG devices,it will be an inevitable trend to automatically identify arrhythmia diseases through algorithms.This paper analyzes and processes the ECG signals of AF and i-avb through data mining technology,and constructs the recognition model of the two diseases with the help of machine learning algorithm,hoping to achieve the early recognition of AF and i-avb diseases,improve the living standards of patients and reduce the burden of medical institutions.In this paper,on the MIT-BIH AF database and the 2018 China first physiological challenge(CPSC2018)database,we effectively combine the P-wave and RR interval features and establish a single recognition model of AF based on the P-wave and RR interval features.At the same time,we will increase the recognition of I-AVB on the basis of traditional AF detection,that is to build an effective multi class recognition model of AF and I-AVB.The main research contents and results are as follows:(1)AF recognition model based on atrioventricular activity.11 short-term features of AF signal were studied and extracted:AF entropy,sample entropy,coefficient sample entropy and 8 statistical features of atrial activity.The Gauss kernel support vector machine is selected as the classifier,and the grid search algorithm is used to optimize the super parameters of the model.The input feature vectors are composed of the short-term features of 11 diagnostic AF signals,and the recognition model of AF disease is constructed.(2)Research on recognition model of AF and I-AVB based on support vector machine.The nonsignificance test is carried out for the studied features,and the qualified features are formed into input feature vectors.The recognition models of AF and I-AVB are built by using the multi classification support vector machine classifier with Gaussian kernel.(3)Research on recognition model of AF and I-AVB based on deep learning.With the help of convolution neural network(CNN),the original ECG signal is used as the input,and a multi input recognition model of AF and I-AVB diseases is built based on CNN.The results show that the performance of AF and I-AVB recognition model based on CNN is better than that of AF and I-AVB recognition model based on machine learning.The accuracy of 87.73%,sensitivity of 87.39%,specificity of 93.91%and value of 87.30%of the model test results on CPSC2018 database are better than most of the published results.At the same time,the ROC curve and AUC value of the model are also good.It shows that the recognition model of AF and I-AVB based on CNN has a certain research value,and can provide a certain reference value for the establishment of the recognition model of a variety of diseases.