Session S82.4

How the Tolerance Threshold "R" Influences Approximate Entropy Analysis of Heart-Rate Variability

P Castiglioni*, M Di Rienzo

Fondazione Don Gnocchi
Milano, Italy

Lack of regularity in physiological time series can be efficiently quantified by the approximate entropy (Apen), an index related to the probability that segments of “m” values which are similar (i.e., closer than the distance “r”) remain similar when the segment length increases to “m+1”. In the past it has been recommended to calculate Apen of heart rate with m=2 and r between 0.10 and 0.25 times the signal standard deviation [1], and actually Apen is now calculated with m=2 and r=0.20 in almost all HRV studies. Recently however it has been shown that when the signal dynamics is faster than the heart rate dynamics, the recommended r values may lead to erroneous conclusions, while the selection of r which maximizes Apen (rM) allows to correctly assess the signal complexity [2]. Aim of this work is to verify: 1) whether rM differs from r=0.20 (the value selected in most HRV studies) also for the slower heart rate dynamics; and 2) whether all the r values within the recommended range (0.10-0.25) provide similar and consistent measures of regularity. We recorded the ECG in 10 subjects (age 21-25 yrs) for 10 minutes, in supine and sitting positions. We calculated Apen of R-R intervals over N=600 beats for m=2 and for r ranging from 0.05 to 1.20. We identified rM (i.e., the r corresponding to the highest Apen), separately in supine and sitting positions. Finally we compared Apen values calculated at r=rM and in the 0.10-0.25 range. We found that rM always felt within the recommended r range; however it differed significantly (p<0.01) from r=0.20, both in supine (rM= 0.18+0.02, m+SD) and in sitting (rM=0.14+0.01) positions. Because of these differences, changes of Apen from supine to sitting were differently quantified: Apen decreased by -6% with r=rM, and by -10% with r=0.20. Comparing the extremes of the recommended range, we even obtained opposite results: in fact, Apen significantly decreased (-12%) with r=0.25, but increased significantly (+9%) with r=0.10. We conclude that the choice of the threshold r is critical even in HRV studies, and that the criterion to select r which maximizes Apen may be helpful also to quantify the heart rate regularity. [1] Pincus SM, Goldberger AL. Am. J. Physiol 266, 1994.[2] Lu S, Chen X, Kanters J, Solomon I, Chon K. Automatic selection of the threshold value r for approximate entropy. IEEE Trans on Biomed Eng, 2008 (in press).

(Abstract Control Number: 215)