Atrial fibrillation, atrial flutter and normal sinus rhythm discrimination by means of blind source separation and spectral parameters extraction

Discrimination between normal sinus rhythm (NSR), atrial fibrillation (AF) and atrial flutter (AFL) using the surface electrocardiogram (ECG) through direct Fourier analysis is not effective because ventricular activity (VA) overlaps in frequency with atrial activity (AA) and, moreover, AA frequency components present considerably lower amplitude than VA. This contribution exploits the assumption that AA and VA can be considered as generated by independent sources of cardioelectric activity. Hence, AA free from spurious VA and other disturbances can be extracted through the application of blind source separation (BSS) techniques. After the blind separation and AA identification process, it was applied a logistic regression over a group of spectral parameters extracted from the power spectral density of the AA signal that was able to discriminate between AF, AFL and NSR with a 100% accuracy considering only two spectral parameters.