Localization of brain alpha activity using independent component analysis infMRI and EEG

This dissertation focuses on the problem of localizing the sources of alpha activity in the human brain using two imaging modalities; functional Magnetic Resonance Imaging (fMRI) and Electroencephalography (EEG), using a novel data processing method called Independent Component Analysis (ICA).; Recently ICA has been introduced to solve the blind source separation problem where unknown statistically independent sources mixed in multi-channel measurements are unmixed only using the statistical properties of data. In this dissertation, we used the recent version of ICA with mixture density model. Its performance was compared against the general versions of ICAs: original ICA and extended ICA. The mixture density ICA, which requires no assumption about the source density, showed the best performance in our simulation study.; In an application of ICA to fMRI, we assume that the sources of alpha activity are spatially independent and the signal intensity in their time courses increase when alpha rhythms are generated. Under these assumptions, ICA directly localized the alpha activity sources in several spatially independent components whose active pixels were concentrated in the parietal, occipital, and frontal lobes. These localizations are in agreement with those of model based approaches such as Linear Correlation and Statistical Parametric Mapping, but are more focal and less noisy at identical significance levels.; EEG data containing alpha activity are also analyzed using the mixture density ICA to isolate multiple alpha components hidden in EEG. The EEG source localization shows that the sources of alpha components are located in the parietal, medial occipital, and the center of the frontal lobes, well-matched with the results of our fMRI study.; One of the regions commonly identified in fMRI and EEG was used as a seed to investigate other functionally connected regions. The result showed that activated regions in both modalities were also connected to each other and thus could be involved in performing a common function.; All proposed methods found a qualitatively good spatial correlation between the locations of the hemodynamic and electrical response to brain alpha activity.