Noninvasive His-bundle electrocardiogram: toward beat-to-beat electromyogram noise removal

EMG noise is a very impeding factor for beat-to-beat detection of His micropotentials in high-resolution ECG. Its reduction was attempted by a two-stage wavelet-domain Wiener filtering procedure. In the first stage, a pilot signal estimate was obtained by averaging a rather small number (∼20) of successive beats synchronized to the R peak. In the second stage, this pilot signal together with EMG noise variance estimate, was used in the wavelet-domain Wiener filtering. The proposed noise suppression technique establishes a compromise between the need of beat-to-beat evaluation of the His potential and the noise-suppression-effective averaging. The algorithm shows adequate performance for records containing both moderate EMG and power-line interference, and for ECGs of sufficiently long P-Q interval (P-wave separated from Q-wave).