Robustness of the Segmented-Beat Modulation Method to Noise

Typically, ECG is corrupted by baseline wander (BW), electrode motion artifact (EM) and muscular artifact (MA). To eliminate them, ECG is usually pre-filtered by application of linear techniques which, however, do not remove in-band components which may limit the ECG clinical usefulness if further processing is not performed. The Segmented-Beat Modulation Method (SBMM) is a template-based filtering technique which segments each cardiac beat into QRS and TUP segments, respectively independent and proportional to heart-rate, and adaptively adjusts each reconstructed beat to its original length by modulating and demodulating the TUP segments. The aim of the present study was to evaluate SBMM robustness to noise by applying it to one synthetic and 18 clinical ECG tracings before and after corruption with BW, EM and MA. Results indicate that, in all cases, clean ECGs are estimated with errors <0.15 mV, typically greater in the QRS than in the TUP segments (0123 µV µV vs 0-25 µV; P<10-5). Moreover, MA little affected ECG estimation, while BW and EM caused higher errors especially in the QRS segment which however remained quite small. Thus, the SBMM resulted to be a filtering technique quite robust to noise.