Modified multiscale entropy in HRV for automatic selection of threshold value r

The multiscale entropy (MSE), a technique, to analyse heart rate variability at different time-scales, requires priori determination of pattern length (m) to be compared and tolerance threshold value (r) to accept the similarity between the patterns. Conventionally recommended value of r, 0.1 to 0.2 times the standard deviation of original time series, found inappropriate for accurate MSE at higher scales. Finding true maximum MSE (MSEmax), by evaluating all values of r from 0 to 1 is a better index of coarse-grained time series complexity, but it is very cumbersome. Automatic selection method of r proposed by Lu et al, has been extended to find MSEmax of RR interval time series for ten healthy subjects with data length N = 900 and m = 2. The burden of finding MSEmax has been averted by the proposed modified multiscale entropy and lead to a more appropriate representation of the complexity than the conventional method.