Master of Science
Mark Joseph Daley
The brain’s underlying functional connectivity has been recently studied using tools offered by graph theory and network theory. Although the primary research focus in this area has so far been mostly on static graphs, the complex and dynamic nature of the brain’s underlying mechanism has initiated the usage of dynamic graphs, providing groundwork for time sensi- tive and finer investigations. Studying the topological reconfiguration of these dynamic graphs is done by exploiting a pool of graph metrics, which describe the network’s characteristics at different scales. However, considering the vast amount of data generated by neuroimaging tools, heavy computation load and limited amount of time and resources, it is vital to refine this pool of metrics to avoid using non-informative and redundant ones. In this study, we use electroencephalographic (EEG) brain signals, taken from recordings in 5 different experimental conditions, to generate the dynamic graphs by moving a sliding win- dow over the time series. Dynamic graphs are produced under various conditions that are a combination of different window sizes, different numbers of shared time points and various frequency bands. Based on each set of these dynamic graphs, time series of 25 graph metrics, and then their pairwise correlation values are computed. This is done to investigate the metric correlations under various circumstances, and to detect the ones that are always present. We conclude by suggesting a set of uniquely informative and orthogonal metrics that is conve- nient to use for further analysis of brain’s functional connectivity
Golestaneh, Mehrsasadat, "Optimizing the Analysis of Electroencephalographic Data by Dynamic Graphs" (2014). Electronic Thesis and Dissertation Repository. 1996.