Ventricular activity residual reduction in remainder ECGs based on short-term autoregressive model interpolation

Considering the inability of existing methods to produce remainder ECGs free from QRS residuals, the present study puts forward a new method for ventricular (QRS) residual detection and reduction in remainder ECGs extracted for the analysis of atrial fibrillation (AF). Autoregressive interpolation (AR) is applied to reduce the amplitude of any QRS residual detected as not negligible by a newly-proposed index, considering the QRS interval as missing, and replacing its samples through interpolation. Performance has been evaluated on a dataset composed of 19 remainder ECGs with AF. Mean (±SD) spectral concentration improved from 56.7±12.8% of the original remainders to 58.1±13.3% of the interpolated ones, while mean (±SD) amplitude of original and processed QRS-T segments was 0.05±0.03 mV. The proposed algorithm was found to improve the quality of the extracted AA remainders without attenuating their mean amplitudes inside the QRST segments.