Multiscale permutation entropy of physiological time series

Time series derived from simpler systems are single scale based and thus can be quantified by using traditional measures of entropy. However, times series derived from physical and biological systems are complex and show structures on multiple spatio-temporal scales. Traditional approaches of entropy based complexity measures fail to account for multiple scales inherent in these time series. Recently multi-scale entropy (MSE) method was introduced, which provide a way to measure complexity over a range of scales. MSE method uses sample entropy, a refinement of approximate entropy to quantify the complexity of time series. Nonstationarity, outliers and artifacts affect the sample entropy values because they change time series standard deviation and therefore, the value of similarity criterion. In this paper, we have used permutation entropy for quantifying the complexity, which is useful in the presence of dynamical and observational noise. We called this modified procedure multiscale permutation entropy (WE). We observed that WE is robust in presence of artifacts and robustly separates pathological and healthy groups.