Search

Browse

• Communities
• Authors
• Year
• Titles
• Subjects

My Space

• Login

Estimation of effective connectivity in resting-state brain networks based on EEG data

Koltsidopoulou Maria-Despoina

Full record

---

Summary

Electroencephalography is widely used to study brain connectivity in healthy individuals as well as changes in connectivity due to various diseases and disorders of brain function. The concept of effective connectivity allows determining connections between different brain regions. These links in the brain are derived from resting-state EEG time series by means of causality analysis methods; the latter aim to quantify cause-and-effect relationships between coupled time series. This thesis uses two different causality analysis methodologies, i.e., Granger causality and information flow rate. Granger causality is based on the linear theory of time series while the information flow rate method is based on the theory of stochastic dynamical systems and the concept of entropy. To the best of our knowledge, the present work is the first attempt to compare the two methods in general and in particular with respect to the calculation of effective brain connectivity.This thesis first investigates the ability of Granger causality and information flow rate to identify causal relationships in synthetic time series derived from simulations. Both methods are applied to simulated data for which the direction of causality is known. Both lead to similar results that are consistent with the expected direction of causality when applied to (a) a stochastic system of two ordinary differential equations and (b) a second-order vector autoregressive system comprising two variables. In addition, the method of random permutations was applied to test the statistical significance of the connections detected by means of information flow. Both methods were also applied to a first-order vector autoregressive system with six variables. The information flow method produces results according to the expected causal relationships, while the Granger causality method detects more causal relationships than those existing in the system. The two methods are then applied to EEG signals obtained from adolescent males in resting-state conditions, in order to map directional interactions between different brain regions. Before applying the causality methods, an exploratory analysis of the EEG data was performed; this includes the estimation of different probability distribution models and empirical variograms as well as spectral analysis. Using the information flow method, we conclude that there is a stronger activity in the occipital lobe of the brain. Using the Granger causality method, we conclude that brain activity focuses on both the anterior and posterior regions of the brain, marked by the transfer of information from one hemisphere to the other. The conclusions above are based on average values over the entire group of 32 healthy individuals. The observed differences in effective brain connectivity based on the two methods require further investigation. Furthermore, the EEG time series were segmented into smaller time windows and reanalyzed using the information flow method. The conclusion drawn from this analysis is that effective connectivity, as calculated by the information flow method, is highly variable in time.

Available Files

Services