Session PB9.3

Noise Effect Analysis in the Non-Invasive Organization Estimation of Atrial Fibrillation

R Alcaraz*, JJ Rieta

Universidad Politecnica de Valencia
Valencia, Spain

The application of nonlinear regularity metrics to physiological signals is a valuable tool because “hidden information” related to underlying mechanisms can be obtained. However, these indices are notably sensitive to noise. To this respect, a 2% noise is serious enough to prevent accurate regularity estimation. In the present work, the noise effect on a previously published method to assess atrial fibrillation (AF) organization via Sample Entropy (SampEn) from ECG recordings is analyzed.
AF organization was estimated by computing SampEn over the atrial activity (AA) signal obtained with a QRST cancellation technique. To evaluate the noise impact on AA regularity, 25 synthetic signals with different organization degrees were generated following a published model. Next, noise coming from real ECG recordings with different energy levels was added to the synthesized AA signals to obtain signal to noise ratios (SNR) of 24, 15, 9 and 3 dB. Finally, SampEn values of the synthetic AA signals contaminated with noise were calculated and compared with those obtained without noise.
Results showed that SampEn, i.e., the AA irregularity, increased with noise, thus hiding the differences between organized and disorganized recordings. Precisely, in the presence of noise, SampEn values were increased, in average, by factors of 1.64, 4.46, 9.46 and 14.23 for SNRs of 24, 15, 9 and 3 dB, respectively. In addition, the difference between the two signals without noise that presented the highest and lowest SampEn values was reduced by factors of 1.07, 1.29, 2.05 and 4.67 for the signals with SNRs of 24, 15, 9, and 3 dB, respectively. These outcomes justify the poor discrimination between spontaneously terminating and non-terminating paroxysmal AF episodes reported by other authors in which SampEn was directly applied to the AA. Moreover, the results are also coherent with the highly improved AF termination prediction reached by applying SampEn to the fundamental waveform associated to the AA, its wavelength being the inverse of the dominant atrial frequency (DAF). As this signal is obtained by applying a selective filtering to the AA centered on the DAF, most part of the undesired contaminating signals are avoided.

(Abstract Control Number: 36)