A new and fast index for the quantification of short range self-similarity in RR time series

Abstract

We propose a new and very fast index (the frequency of sign changes of the mirrored differences or fscmd) with good correlation with the short range scaling exponent (  Ｒ) estimated among scales 4 to 16 of the DFA. fscmd computes the relative number of sign changes of the difference of the RR time series minus their corresponding reversed RR time series after a moving average detrending procedure is applied using a window of 30 samples. Linear regression results with simulated time series with Fractional Brownian Noise and with actual time series using the Fantasia (FT), Normal Sinus Rhythm RR time series (NSR) and Congestive Heart Failure RR interval (CHF) databases after artifact correction show good agreement between fscmd and  Ｒ. Finally, Mann-Whitney Rank Sum tests applied to  Ｒ and fscmd when comparing NSR and CHF databases show very significant differences (p<0.001) between groups for both indices.