Problems of analysis of electroencephalograms by methods of nonlinear dynamics

In this paper, the analysis of the problems of processing electroencephalograms by non-linear dynamics methods is made. It is shown that the results obtained by different methods, including machine/deep learning methods, neural networks allow classifying an epileptic sei-zure based on EEG data with an accuracy of 94% or higher. The problems that occur when performing these studies are indicated. The problems of analyzing data taken from different databases that arise when using nonlinear dynamics methods are considered. Their compari-son and clustering are carried out according to the JRQA indicators of EEG data analysis and Hurst indices. However, it is currently not possible to identify an analytical dependence in the direction of predicting an epileptic seizure.