The Study Of Atrial Fibrillation Recognition Based On Support Vector Machine And The Design And Implementation Of Common Arrhythmia Monitoring System Model

Whether it is a portable single-lead electrocardiogram(ECG)acquisition device or a multi-lead ECG acquisition device,the collected ECG signals may include noises due to breathing,movement,lead-off,etc.The appearance of ECG artifacts will lead to distortion of ECG parameters(such as RR interval,QT segment,etc.),causing misdiagnosis or missed diagnosis based on ECG.Severe ECG artifacts may mask the true signal and make the masked ECG signals lose values of diagnosing and monitoring cardiac disease.Therefore,the quality of ECG can be improved effectively by identifying and removing ECG artifacts.Atrial fibrillation(AF)is a common arrhythmia disease.With the acceleration of urbanization and aging of society,its incidence is increasing.An efficient and accurate automatic identification of atrial fibrillation method is the technical guarantee for large-scale atrial fibrillation monitoring and management.To this end,this paper uses support vector machine(SVM)technology to carry out researches on the identification methods of ECG artifact and atrial fibrillation.Based on this,a common arrhythmia monitoring system model is developed to improve the early identification and monitoring of multiple arrhythmia based ECG signals,and provide technical support in real-time monitoring of health status at any time and anywhere.The specific research contents are as follows:(1)ECG artifact identification method with multi-feature parameters based on SVM.This method firstly extracts five signal quality indicators(SQI),including the ratio of the first principal component of the ECG,the standard deviation of the R-wave amplitude,the correlation coefficient of template matching,the energy ratio of the QRS wave and the sample entropy.Then,using five SQI,the classification and recognition model of ECG artifacts and signals is constructed based on SVM of Grid parameter optimization.Next,a validation dataset is built through merging the ECG artifacts and ECG signal from the 4 ECG databases: PhysioNet Challenge 2011,PhysioNet Challenge 2017,MIT-BIH arrhythmia Database,Noise Stress Test Database.Finally,the 10-fold cross-validation is applied to verify the performances of the proposed model.As a result of the experiment,the model has the sensitivity,specificity,positive predictive and accuracy of 98.33%,98.14%,98.04% and 98.24%,respectively,and has strong ability of detecting and identifying ECG artifacts.(2)AF identification method based on unbalanced multi-classification SVM.This method first extracted 134 candidate features from existing researches and removed 24 features related to P-waves because P-wave is difficult to be accurately located.Then,the correlation analysis was made among the remaining 110 candidate features.An effective feature set is formed after removing the features of high redundancy with the correlation coefficient of > 0.9 and the features of high complexity from remaining 110 candidate features.Next,an unbalanced four-class SVM classifier was designed based on the distribution of different types of ECG data and it was combined with the effective feature set to detect four types of ECG signals,including AF,other arrhythmia,artifacts and normals.Finally,the real ECG data provided by the PhysioNet/Computing in Cardiology Challenge 2017 confirmed that the proposed method had a overall good performance compared with five other related methods.Also,the data from MIT Arrhythmia Database and the MIT Atrial Fibrillation Database confirmed the robustness of proposed method with AF detection score of > 0.97 and with the scores of > 0.9 in other arrhythmia,artifacts and normals.The proposed method has a good application prospect in aided diagnosis,monitoring and management of AF.(3)Design and implementation of AF and non-AF arrhythmia monitoring system model.According to the real needs of AF and other arrhythmia monitoring and management,a model of AF and non-AF arrhythmia monitoring system based on ECG signal was designed.The ECG signal acquisition model is used to simulate the data acquisition process of the real system.The ECG signal wireless transmission model is used to simulate the transmission process between the data.The ECG signal intelligent processing model is used to simulate the signal processing process of the real system,and the ECG signal display model is used to simulate the display of the discriminant results by the real system.The test results show that the model has comprehensive functions,real-time and operability,friendly interface and convenient use,and provides the theoretical and technical support for practical application.