Study on the alternatives to reduce high-frequency noise from invasive recordings of atrial fibrillation

Abstract

Catheter ablation (CA) is nowadays the first-line therapy for the treatment of atrial fibrillation (AF). This approach relies heavily on cardiac mapping systems, which provide intra-cardiac electrograms (EGMs) to study the heart's electrical activity. To get precise clinical information from these signals, the elimination of noise and nuisance interferences, as well as the preservation of their original waveform, are key aspects to treat carefully. However, how they should be preprocessed to remove successfully high-frequency noise has only received limited attention to date. Hence, in this study the most commonly bandpass filtering applied to this recording is compared with two advanced denoising methods, based on Wavelet Transform (WT) and Empirical Mode Decomposition (EMD), on a set of 150 synthesized bipolar EGMs with different levels of noise. The resulting signals have been contrasted with the original ones in terms of a signed correlation index (SCI), computed both from time and frequency domains. Thus, notably more significant alterations in the original waveform have been observed for the regular filtering than for the WT-based approach. As a conclusion, this method is a more interesting option than the regular filtering to preserve the morphology of the bipolar EGM and obtain more accurate electrophysiological information.