Feature extraction and analysis of atrial activity from surface electrocardiogram with atrial fibrillation

Objective A new method of blind source separation is proposed to extract atrial activity signals from surface electrocardiogram and analyze the features of atrial activity during atrial fibrillation(AF) episodes.Metheds The wavelet analysis was used to implement the sparse representation of ECG.Those three fundamental requirements which must be satisfied in independent component analysis(ICA) theory were verified in wavelet domain and a mathematical model of ICA in wavelet domain was presented.Results The atrial activity signals were extracted by the ICA model in wavelet domain.Some metrics such as the P wave dispersion(Pdis),frequencies spectrum and AF cycle length(AFCL) which had close relation with AF were discussed in details.And features of different kinds of atrial fibrillation were compared.Conclusion These new non-invasive approaches have the ability to extract atrial activity signals during AF episodes,which have great importance to the mechanism research and clinical treatment of AF.