A comparative stationarity analysis of EEG signals


RASOULZADEH V., ERKUŞ E. C., YOĞURT T. A., ULUSOY İ., Zergeroglu S. A.

ANNALS OF OPERATIONS RESEARCH, vol.258, no.1, pp.133-157, 2017 (SCI-Expanded) identifier identifier

  • Publication Type: Article / Article
  • Volume: 258 Issue: 1
  • Publication Date: 2017
  • Doi Number: 10.1007/s10479-016-2187-3
  • Journal Name: ANNALS OF OPERATIONS RESEARCH
  • Journal Indexes: Science Citation Index Expanded (SCI-EXPANDED), Scopus
  • Page Numbers: pp.133-157
  • Keywords: EEG, Stationary, VEP, Fractal, Hilbert-Huang transform, Statistical analysis, INTERHEMISPHERIC-TRANSFER TIME, EMPIRICAL MODE DECOMPOSITION, VISUAL-EVOKED-POTENTIALS, SERIES, VEP
  • Middle East Technical University Affiliated: Yes

Abstract

While developing models of brain functioning by using time series data, the stationary interval of the time series should be used to model the corresponding state of the brain. Here it is assumed that at the borders of stationarity, brain changes its state where a state is considered as a group of brain regions working together. If the whole nonstationary time series is used, many different brain states could be included in one model. However, it is very hard to decide the stationary intervals for such a complicated system as brain. There are some methods, which have proved their performances based on manually constructed synthetic data, in the literature. Only very few results with EEG data have been presented. But, there is usually no ground truth accompanying the data to make an evaluation. In this study, suitable approaches for stationary analysis were applied on visual evoked potentials (VEP) where we can approximately know the possible stationary intervals due to the properties of the experiment during which the data was recorded. Experts designed the experiment and marked the possible borders of the intervals carefully. Parameters of the methods were set automatically. We compared the manually marked intervals with the intervals detected automatically by the applied methods and evaluated the methods in terms of their performances of estimating the stationary intervals of EEG signals.