QRST cancellation in ECG signals during atrial fibrillation: Zero-padding versus time alignment

In non-invasive atrial fibrillation (AF) ECG analysis, it is common to remove the QRST complex components by a time-aligned average over several ECG cycles. This paper proposes another alternative using zero-padding of the QRST segments which introduces less illusive high frequency content into the residual signal. The framework of this study follows three main steps: i) characterize the sub-segments of the ECG cycles which are associated with the AF episode, ii) implement a procedure for detecting QRS complexes and T waves using robust ECG signal processing techniques and, iii) cancel the detected QRST complexes in the abnormal ECG signal via two techniques which include attenuating the QRST segment by time aligned averaging and zero-padding. The Pan Tompkins' algorithm was utilized to detect Q, R and S peaks. Then, the Zhang's algorithm was utilized to find the onset and offset of each T wave. Subsequently, the QRST segment was determined in each period of the ECG signal. Then, two different approaches were applied to remove the QRST segments from each period and make the signal ready for AF analysis. In order to compare the efficiency of the QRST cancellation approaches, mean spectral power of the residual signal at high frequency bands (upper 12 Hz) was used. The results imply the superiority of the time-aligned averaging method for the application of AF analysis.