Atrial electrical activity extraction for atrial fibrillation assessment

The work is dedicated to noninvasive research of atrial electrical activity by its extraction from the surface electrocardiogram recordings and spectral and time-frequency analysis for fibrillatory frequency detection and tracking. Atrial electrical activity and ventricular activity were separated from real multichannel recordings with symptoms of atrial fibrillation by means of independent component analysis. To identify the source that corresponds to atrial electrical activity the frequency spectrums and coefficients of skewness and kurtosis were determined. Also atrial activity extraction was performed by average beat subtraction method. The comparison of obtained results was carried out. Spectral analysis by Fast Fourier Transform and Multiple Signal Classification methods as well as time-frequency analysis were used to find and track dominant atrial fibrillation frequency.