ISBN 9783843905602

978-3-8439-0560-2, Reihe Ingenieurwissenschaften

Matthias Wacker
Dynamic phase analysis of MEG and EEG signals

100 Seiten, Dissertation Technische Universität Ilmenau (2012), Hardcover, A5

Zusammenfassung / Abstract

A photic driving experiment is a good scenario to study the dynamic properties of brain oscillations. By a repetitive visual stimulus, the brain operates in a rhythmic mode, where phase coordinations in the excited neuronal oscillators can be observed. Due to the variety of expectable synchronization effects, photic driving can be seen as one of the best scenarios to put phase analysis methods to the test. The presented investigations revealed three major limitations of state-of-the-art methods, which required the development of new analysis techniques.

First, the entrainment of the alpha oscillation to the stimulus has to be quantified by instantaneous frequency estimation. As the activity is intermittent and phase jumps occur, the estimation is highly error prone. A new algorithm is proposed that robustly estimates the instantaneous frequency as the time-frequency trajectory of maximum energy. An efficient energy extraction is achieved by spline-based chirp atoms to speed up the optimization algorithm.

Second, the predefinition of an optimal time-frequency resolution is a crucial point for the analysis. State-of-the-art methods for frequency-selective phase extraction enforce a certain time-frequency appearance, which depends on the chosen parameters of the method. To maintain the original time-frequency characteristic of the signal, a new analysis method is introduced. It decomposes the signal into atoms which are analyzed with Gabor Transforms, where the analysis time-window matches the atom envelope. This results in the first data-adaptive, frequency-selective phase extraction and makes the time-frequency appearance an interpretable feature.

Third, possible synchronizations between the alpha oscillation and the gamma band activity are analyzed with the n:m phase synchronization index. It measures the consistency of the generalized phase difference between two frequencies among all trials. State-of-the-art results for certain frequency combinations contain a very high noise level so that even strong synchronizations can be hidden. A new frequency-tiling approach is presented that avoids large n and m values and with that, reduces the noise level. It focuses on the identification of stable synchronizations while circumventing unstable effects and thus increases the overall reliability and usability of the synchronization analysis.