Analysis of scaling behaviour of ECG signal during atrial fibrillation

In this study, surface ECG signals recoded during Atrial Fibrillation (AF) episodes have been investigated to assess the presence of scaling behaviors. The ECG signals consisted of one minutes recordings coming from the 2004 Computers in Cardiology Challenge database. After having removed the ventricular related activity, the residual ECG signal (rECG) was analyzed using Detrended Fluctuation Analysis (DFA). We documented the existence of two scaling regions in the signals. Over long time scales (n >= 32, approximate to 234ms) the rECG fluctuations resembled a white or slightly persistent noise (scaling A=0.64 +/- 0.17), whereas over short scales (n <= 16, approximate to 125ms) slowly decaying antipersistent correlations were evident (scaling A=1.20 +/- 0.20). In addition, we searched for possible differences in the scaling exponents between terminating (T) and nonterminating (N) AF episodes. We observed that both short-term (A1: 1.29 +/- 0.15 for T vs. 1.12 +/- 0.18 for N, p < 0.005) and long-term (A2: 0.54 +/- 0.16 vs. 0.70 +/- 0.14, p < 0.005) scaling exponents were significantly different (t-test).