New methods for signal-adaptive, time-variant analysis of phase properties and directed interactions of and between EEG/MEG oscillations

• Applicants: Dr. Lutz Leistritz; Professor Dr. Herbert Witte
• Subject Area: Epidemiology and Medical Biometry/Statistics
• Term: from 2012 to 2017
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 231138331

The advantage of signal-adaptive, time-variant analysis methods is that (e.g.) time-frequency properties can be analysed in a signal-driven and signal-optimal manner. The proposed methodological approaches are composed of two processing units. In the first unit, signals are decomposed into signal-atoms or signal components. The resulting components are subsequently processed (directly or via a suitable elimination scheme) by a specific analysis method in the second unit. According to the envisaged signal property, both processing units form a holistically optimized analysis concept. Phase properties and directed interactions (effective connectivity) of and between EEG/MEG oscillations are of particular interest. For an optimal time-frequency analysis of the instantaneous phase the Matched Gabor Transform, which was introduced by us, will be generalized. Additionally, the methodology of the Hilbert-Huang Transform (HHT) will be improved and advanced to enable multi-trial as well as multi-channel analyses. The concept of the Granger causality will be expanded by elimination schemes for frequency components, which are extracted by filtering, to optimize frequency-selective connectivity analysis. Furthermore, we aimed at a replacement of filtering by Empirical Mode Decomposition (part of the HHT) and a decomposition of signals in ARMA(2,1) components. This would lead to the possibility to perform signal-adaptive, frequency-selective connectivity analyses. All new methods will be tested by simulated and clinical data and a critical comparison with the results provided by other methods is planned.

• DFG Programme: Research Grants