Several insights into the preprocessing of electrograms in atrial fibrillation for dominant frequency analysis

Abstract

BackgroundDominant frequency (DF) analysis of atrial electrograms has become an important method in characterizing atrial fibrillation (AF). As a classic method, Botteron’s approach is widely used in the preprocessing of frequency analysis during AF. It includes three steps: (1) band-pass filtering at 40–250 Hz, (2) absolute value, and (3) low-pass filtering at 20 Hz. This paper aims to expound the necessity and adjustability of each step.Methods and resultsUnipolar epicardial mapping signals were recorded during AF from eight mongrel dogs with cholinergic AF model. Episodes of these data were randomly selected to evaluate the impact of different pass bands and the necessity of low-pass filtering with 20 Hz cutoff frequency. Each episode of AF signal is 5 s long with a sampling rate of 2 kHz. Simulated electrograms were adopted to discuss the role of taking absolute value. Furthermore, direct spectral analysis method (FFT et al.) is compared with Botteron’s preprocessing approach. According to our statistical analysis, the pass band of 40–250 Hz was not the best, while 20–100 Hz presented the high accuracy rate of DF. From the comparing result of direct FFT without Botteron’s approach we deduced that the rectification of absolute value was meaningful for the fundamental atrial frequency. The final step, 20 Hz low-pass filter can completely be omitted in DF analysis. In consideration of the demand for real-time distribution of DF in clinical or experimental situations, down-sampling method and the impact of ventricular artifacts on DF was also discussed.ConclusionIn the actual application of the three preprocessing steps, the pass band selection of band-pass filter can be adjusted and the rectification of taking absolute value is important. Nevertheless, the final step of 20 Hz low-pass filter is totally unnecessary. In real-time signal processing situations, taking down-sampling method and ignoring the ventricular artifacts can also have high performance in DF analysis of atrial electrograms.

Systematic differences of non-invasive dominant frequency estimation compared to invasive dominant frequency estimation in atrial fibrillation

“…We investigated the effect of altering pre-processing steps on extracting HDFs as proposed by Li and colleagues [ 25 ], namely filtering the VEGMs between 20 and 100 Hz and rectifying these filtered signals. We examined the effect of estimating the AF spectral content using Welch's method and the FFT method.…”

“…An elaborate intuition on how to derive and interpret HDFs from endocardial data has been presented by Ng and Goldberger in a series of publications [ 20 , 23 , 24 ]. Further recommendations on deriving frequency spectra using Welch periodograms [ 4 ] and optimal pre-processing steps for unipolar electrograms were recently presented [ 25 ]. Following these recommendations, it was shown that dominant frequency (DF) analysis could be performed reliably on both unipolar and bipolar endocardial signals [ 23 ].…”

Classification of Atrial Fibrillation Recurrence Based on a Convolution Neural Network With SVM Architecture

“…Secondly, we would get the absolute value of the outputs from previous procedure, and finally the signal would pass a third-order Butterworth filter with a 20 Hz cutoff. After such processing, the ventricular activity can be greatly suppressed [22].…”

Leads Selection of Body Surface Potential Mapping During Atrial Fibrillation: A Sequential Selection Based on Adapted Botteron’s Approach

“…But it was firstly proposed as the assessment of the spatial organization by means of a cross-correlation approach [21]. In our previous study [18], [22], we applied Botteron's approach to BSPM signals as a rhythm analysis tool. In this study, we modified the Botteron's approach to apply it to the derived AF signals in order to obtain rhythm-related indicator waveform for further lead selection.…”

“…Based on the spectral content of derived AF signals and the theoretical basis of Botteron's approach [22], each step of Botteron's approach was modified. The signals were firstly band-pass filtered using a digital, zero-phase, third-order Butterworth filter with the range of 4-60 Hz instead of 40-250Hz.…”