Novel way to investigate evolution of children refractory epilepsy by complexity metrics in massive information.

Epilepsy demands a major burden at global levels. Worldwide, about 1% of people suffer epilepsy and 30% of them (0.3%) are anticonvulsants resistant. Among them, some children epilepsies are peculiarly difficult to deal with as Doose syndrome (DS). Doose syndrome is a very complicated type of children cryptogenic refractory epilepsy (CCRE) which is traditionally studied by analysis of complex electrencephalograms (EEG) by neurologists. CCRE are affections which evolve in a course of many years and customarily, questions such as on which year was the kid healthiest (less seizures) and on which region of the brain (channel) the affection has been progressing more negatively are very difficult or even impossible to answer as a result of the quantity of EEG recorded through the patient's life. These questions can now be answered by the application of entropies to massive information contained in many EEG. CCRE can not always be cured and have not been investigated from a mathematical viewpoint as far as we are concerned. In this work, a set of 80 time series (distributed equally in four yearly recorded EEG) is studied in order to support pediatrician neurologists to understand better the evolution of this syndrome in the long term. Our contribution is to support multichannel long term analysis of CCRE by observing simple entropy plots instead of studying long rolls of traditional EEG graphs. A comparative analysis among aproximate entropy, sample entropy, our versions of multiscale entropy (MSE) and composite multiscale entropy revealed that our refined MSE was the most convenient complexity measure to describe DS. Additionally, a new entropy parameter is proposed and is referred to as bivariate MSE (BMSE). Such BMSE will provide graphical information in much longer term than MSE.