Permutation relative entropy in quantifying time irreversibility: loss of temporal asymmetries in epileptic EEG

Permutation relative entropy is proposed to quantify time irreversibility in our nonlinear dynamics analysis of electroencephalogram (EEG). Ordinal patterns in multi-dimension phase space of time series are symbolized, and the probabilistic divergences of all symmetric ordinal pairs are measured by relative entropy as time irreversibility. Analyzing multi-channel EEG from 18 healthy volunteers and 18 epileptic patients (all in their seize-free intervals), the derived relative entropy of symmetrical ordinal patterns shows advantages to other time irreversible parameters and significantly distinguish two kinds of brain electric signals (the epileptic have lower temporal asymmetries than the healthy). Test results prove that it is effective to quantity time irreversibility by measuring probabilistic divergence of symmetrical ordinal patterns and validate our hypothesis that epilepsy has lasting impacts on brain’ nonlinear dynamics, leading to a decline in brain signals directional asymmetry or time irreversibility, and the losing temporal asymmetries stemming from our findings may contribute to the preclinical diagnosis of epilepsy.