gms | German Medical Science

GMDS 2013: 58. Jahrestagung der Deutschen Gesellschaft für Medizinische Informatik, Biometrie und Epidemiologie e. V. (GMDS)

Deutsche Gesellschaft für Medizinische Informatik, Biometrie und Epidemiologie

01. - 05.09.2013, Lübeck

New Methodical Developments for Dynamic Analysis in Time-Frequency Domain

Meeting Abstract

  • Galina Ivanova - Humboldt-Universität zu Berlin, Berlin, DE
  • Katrin Pauen - Humboldt-Universität zu Berlin, Berlin, DE
  • René Heideklang - Humboldt-Universität zu Berlin, Berlin, DE
  • Irina Katsarska - Humboldt-Universität zu Berlin, Berlin, DE

GMDS 2013. 58. Jahrestagung der Deutschen Gesellschaft für Medizinische Informatik, Biometrie und Epidemiologie e.V. (GMDS). Lübeck, 01.-05.09.2013. Düsseldorf: German Medical Science GMS Publishing House; 2013. DocAbstr.337

doi: 10.3205/13gmds268, urn:nbn:de:0183-13gmds2687

Veröffentlicht: 27. August 2013

© 2013 Ivanova et al.
Dieser Artikel ist ein Open Access-Artikel und steht unter den Creative Commons Lizenzbedingungen ( Er darf vervielfältigt, verbreitet und öffentlich zugänglich gemacht werden, vorausgesetzt dass Autor und Quelle genannt werden.



Introduction and Objective: The acquirement and analysis of diverse brain signals is a standard tool in neuroscience, psychology and other related disciplines, whereby the analysis of cognitive correlates is becoming increasingly important. Two main experimental design concepts are applied in the praxis: the acquirement of both single-trial and of multi-trial data sets. In consequence different signal analysis procedures need to be applied dependent on the signal characteristics e.g. signal-to-noise-ratio or variability of the relevant research context components. We introduced recently three different techniques for analysis of single- and multi-trial signals in the time-frequency domain. Here we demonstrate the application of these methods and of their combination for the investigation of cognitive brain signals. Our intention is to explore in particular how the extraction of information could be improved in comparison to conventional procedures. Furthermore we discuss the applicability to diverse biomedical and technical signals.

Methods and Results: The first methodology was derived from the development by N.E. Hunag for NASA, a combination of intrinsic mode functions and subsequent Hilbert transform. One of the main difficulties in the algorithmic process is to find the correspondence between the functions resulting from the decomposition of different signals. The reasons for this are the inherent heuristic and the difference in the variations in the number of the estimated functions. Our adaptation of this technology allows spatial and multi-trial signals analysis. A very good interpretability of the dynamics in time, frequency and space is achievable in the case of a successful decomposition. The second group of methods introduced by our group is for the analysis of phase relationships in the circular domain and includes several techniques based on circular circular correlations. These methods allow more appropriate examination of phase-time series, because the mathematical foundations were developed especially for the calculation of angles. Furthermore the new measures focus on the explanation of synchronization and couplings not only in bivariate but also in multivariate cases. Therefore the exchange of information between more than two brain areas is detectible by only a single measure. The third methodology is the smooth natural Gaussian extension (snaGe). This technique allows an adaptive modeling of the time-frequency distributions of non-stationary signals and the estimation of interpretable parameter sets describing the data. Various applications of this method are provided e.g. the selection of signals having comparable time-frequency dynamics, non-linear filtering, data-compression and others. As mentioned above a combination of the method e.g. circular analysis or a snaGe-modeling of intrinsic mode components are considered and will be discussed.

Discussion: In conclusion the developed methods represent promising tools for analysis of single- and multi-trial data. New insights in the temporal-, frequency-, spatial- and in the coupling signal dynamics could contribute decisively to a better understanding of the underlying information processing and the process generating structures. Several applications of the methods to a high number of biological and technical time-series are easily possible and are planned for the near future.