Research On Automatic Detection Method Of Atrial Fibrillation For Mobile Health Care

Atrial fibrillation(AF)is the most common type of arrhythmia in cardiovascular disease.Early detection of atrial fibrillation has important implications for the prevention and treatment of heart disease.As the population ages,the incidence of atrial fibrillation increases,and the medical resources are relatively limited.Mobile health care has many advantages in the detection of atrial fibrillation and has attracted more and more attention.Physiological signal acquisition devices in mobile medical system contain limited number of electrical leads,usually only including single leads,which brings challenges to the accurate detection of atrial fibrillation.Aiming to solve problems mentioned above,in this paper research on automatic detection based on single lead short-term ECG signal for AF is presented.Major contributions include:(1)An AF automatic detection algorithm based on the atrial activity signals and RR interval was proposed.Having analyzed the traditional AF detection method’s deficiency based on K-S test,this paper introduces an improved scheme of “fuzzy area” division.In order to make full use of the rhythm information contained in the short-term ECG signals,the time and frequency domain characteristics of atrial electrical signals and RR interval characteristics were fused to detect AF.Aiming at different AF features,a two-stage feature extraction method is proposed to improve the accuracy of AF detection.(2)A combined PVC recognition method for AF detection was proposed.Having analyzed the characteristics of premature ventricular contraction(PVC)and its effect on AF detection,an improved method for PVC recognition is proposed.Based on the analysis of AA signals and RR interval,an ectopic beat removal strategy based on this method is introduced.Finally,the AF automatic detection algorithm combined with PVC recognition is realized,which further improved the accuracy of AF detection.Results from the MIT-BIH atrial fibrillation database showed a sensitivity(SE)of 96.1% and specificity(SP)of 95.5% when the detection window length was set to30 beats,and the proposed algorithm based on single lead short-time ECG data had a better performance compared with other algorithms.