Effects of the ECG sampling frequency on the multiscale entropy of heart rate variability

It is known that the spectral analysis of heart rate variability requires an ECG sampling frequency Fs>1 00 Hz with parabolic interpolation to refine the R peak if Fs<250Hz. By contrast, the effects of quantization errors in Multiscale Entropy (MSE) analysis due to low Fs have never been evaluated systematically. Our aim is thus to describe the effects of low Fs and parabolic interpolation on MSE. We considered 21 ECG recordings of 10' duration sampled at 500Hz (reference). We decimated the ECG to simulate Fs between 250 and 50Hz, we extracted the tachograms without and with parabolic interpolation and estimated MSE at scales between 1 beat (=SampEn) and 50 beats. The estimates were expressed as the percentage of the reference and the error was quantified by the interquartile range (IQR) of their distribution. SampEn showed high sensitivity to Fs with IQR > 1 0% at 250Hz and >16% at 167Hz; however, the parabolic interpolation dramatically decreased the IQR below 2% up to Fs=71Hz. The MSE estimates at larger scales were less sensitive to Fs with IQR=2% even at Fs=50Hz. Thus the ECG sampling rate is more critical for SampEn than for MSE at larger scales and interpolation procedures are required when Fs<250Hz.