Wavelet Entropy Automatically Detects Episodes of Atrial Fibrillation from Single-Lead Electrocardiograms

Abstract

This work introduces for the first time the application of wavelet entropy (WE) to detect episodes of the most common cardiac arrhythmia, atrial fibrillation (AF), automatically from the electrocardiogram (ECG). Given that AF is often asymptomatic and usually presents very brief initial episodes, its early automatic detection is clinically relevant to improve AF treatment and prevent risks for the patients. After discarding noisy TQ intervals from the ECG, the WE has been computed over the median TQ segment obtained from the 10 previous noise-free beats under study. In this way, the P-waves or the fibrillatory waves present in the recording were highlighted or attenuated, respectively, thus enabling the patient's rhythm identification (sinus rhythm or AF). Results provided a discriminant ability of about 95%, which is comparable to previous works. However, in contrast to most of them, which are mainly based on quantifying RR series variability, the proposed algorithm is able to deal with patients under rate-control therapy or with a reduced heart rate variability during AF. Additionally, it also presents interesting properties, such as the lowest delay in detecting AF or sinus rhythm, the ability to detect episodes as brief as five beats in length or its integration facilities under real-time beat-by-beat ECG monitoring systems. Consequently, this tool may help clinicians in the automatic detection of a wide variety of AF episodes, thus gaining further knowledge about the mechanisms initiating this arrhythmia.

Combined Nonlinear Analysis of Atrial and Ventricular Series for Automated Screening of Atrial Fibrillation

“…Precisely, the most recent works are putting special emphasis on palliating this issue [23,[25][26][27][28][29][30]. In fact, some authors have proposed the use of information from RR intervals series irregularity in combination with features obtained from the atrial activity (AA), that is, the P-orwaves [23,28,29].…”

Robust Feature Extraction from Noisy ECG for Atrial Fibrillation Detection

“…We ran several experiments on different number of segments, m ∈ {5, 6, ..., 30} and different sampling frequencies f sampling ∈ {75, 150, 300}. We started with m = 5 as the minimum number of RR segments used in [8]. For these experiments we use all the features generated by FE.…”

Classification of Atrial Fibrillation Using Multidisciplinary Features and Gradient Boosting

“…A more advanced set of full waveform features was generated from the stationary wavelet transform decomposition of the full waveform [6]. The transformation was done using the PyWavelets toolbox and the Daubechies 4 (db4) mother wavelet.…”

“…On each set of coefficients, three max/mean power spectral density ratios were calculated for the following frequency bands: 3 -10 Hz, 10 -30 Hz, and 30 -45 Hz. Additionally, the log entropy [6] and Higuchi fractal dimension (PyEEG toolbox) were calculated.…”