Main research topics:

Granger causality is a basic concept in the analysis of multivariate time series and regards causal relationships among the variables of the underlying dynamical system or stochastic process. In many applications, including electroencephalograms (EEG), the estimated causal effects can give insight into the system structure. Recently, it has been suggested to form networks having connections formed by the estimated causal effects, study them using network measures, and further detect structural changes in the network time evolution.

This  work lies in this framework and focuses on the following:

a) The statistical assessment, using resampling and significance testing, of the most known linear (in the time and frequency domain) and nonlinear (including information measures) Granger causality measures, including measures of direct causal effects in the presence of confounding variables.

b) The investigation of the effectiveness of the network approach in detecting structural changes.

c) The classification of different system regimes on the basis of causality and network measures.

This is a novel approach and involves feature selection in order to determine the most relevant measures. The analysis will be performed via Monte Carlo simulations on properly designed systems. In parallel, the proposed analysis will be applied in two EEG case studies:

i) The transcranial magnetic stimulation (TMS) combined with EEG: the connectivity of brain areas of human epileptic subjects will be estimated before and after TMS at normal activity and during epileptic discharges.

ii) The discrimination of preictal states on the basis of brain connectivity estimated on human epileptic EEG. The research group has active and high quality research record on both the methodological and neurophysiological part of the project.

  1.  Participation in conferences realative to the project of  ARISTEIA II 

1.      XXXIV Dynamics Days Europe 2014 -07 - 13/09/2014 - BAYREUTH: oral SIGGIRIDOU, KUGIUMTZIS
2.	NOLTA 2014, 14 - 18/09/2014 – LUZERN: oral KUGIUMTZIS
3.      9th Panhellenic Congress of Epilepsy, 17- 20/10/2014 – Athens:  Poster KUGIUMTZIS
4.	28th Panhellenic Statistics Conference, 15- 18/04/2015, Athens: oral ΚUGIUMTZIS, SIGGIRIDOU, TSIMPIRIS 
5.      37th EMBS, 24/08/2015 - 30/08/2015, MILAN: oral KUGIUMTZIS 
6.      Dynamics Days Europe 2015, 06 - 11/09/2015, EXETER: oral KUGIUMTZIS, SIGGIRIDOU 
7.      15th European Congress on Clinical Neurophysiology, 30/09-03/10/2015, BRNO: oral KIMISKIDIS 
8.      FFRM 2015, 07/10/2015 - 10/10/2015, Thessaloniki : oral KOUTLIS, TSIMPIRIS, SIGGIRIDOU

 

Other research topics:

  • Evaluation and development of measures for univariate time series and application to multi-channel preictal EEG records.
  • Detection of changes in preictal EEG records, classification of preictal states: use of measures for univariate time series and data mining techniques.
  • Development of software for the implementation of measures of time series analysis, as well as data mining techniques for feature-based clustering, using the measures as features.
  • Evaluation and development of causality measures for bivariate and multivariate time series and application to multi-channel preictal EEG records.
  • Evaluation and development of complex networks from multivariate time series and application to multi-channel preictal EEG records.
  • Feature selection on measures of univariate and multivariate time series analysis and application to classification tasks on multi-channel EEG records.

Methodology:    

  • Study of the significance of measures for coupling and causality in multivariate time series.
  • Development of causality measures on multivariate time series when the number of variables is large.
  • Investigation of methods for dimension reduction in the analysis of multivariate time series.
  • Study of clustering approaches based on time series features (feature-based clustering).
  • Study of complex networks from multivariate time series using correlation and causality measures

Applications to physiology:    

  • Development of a data base of preictal and ictal EEG records, clinical description and results of statistical analysis of the records.

  • Use of techniques of analysis of univariate, bivariate and multivariate time series on EEG records in conditions of transcranial magnetic stimulation.

  • Use of techniques of analysis of univariate, bivariate and multivariate time series on epileptic EEG records.