Adaptive singular value QRST cancellation for the analysis of short single lead atrial fibrillation electrocardiograms

Atrial fibrillation (AF) is the most commonly diagnosed sustained supraventricular arrhythmia in clinical practice and is characterized by uncoordinated atrial activation. The proper analysis and characterization of AF from surface ECG recordings requires to cancel out ventricular activity (i.e. QRS-T complexes). Some powerful methods exploit the spatial diversity of multi-lead ECG, however, their performance is seriously reduced in single-lead environments. For this latter case, techniques based on averaged beat subtraction (ABS) are the most widely used. However, these methods are very sensitive to QRS-T wave variations, thus, a high quality QRS-T cancellation template may be difficult to obtain when only short length recordings are available. To overcome these difficulties, a new QRS-T cancellation method based on singular value decomposition (SVD) of each single beat is presented. This methodology was tested and validated using a significative database with simulated and real AF recordings. The results showed that SVD is able to obtain a very accurate ventricular activity (VA) representation, thus providing high quality atrial activity (AA) extraction in short and single-lead AF recordings. Therefore, the inherent limitations in ABS techniques due to variations in the QRS-T shape could be avoided.