Independent Component Analysis For Higher Brain Function

The dissertation is devoted to some models and methods about independent component analysis (ICA) and its application to the analysis of functional magnetic resonance imaging (fMRI) data.Independent component analysis is a powerful tool for data processing, which was first proposed in 1980s. Some important results in theory and application have been obtained, it has been widely used in various fields. Some objective functions of ICA, such as maximum likelihood estimation, mutual information minimization, infomax and non-gaussian, are summarized. Then two famous algorithm (fastICA and infomax ) of ICA are introduced. Finally the application of ICA in biomedical signals processing is discussed.Functional magnetic resonance imaging (fMRI) is a no-invasion tool for the research of brain, by which the change of the brain in all kinds of behavior can be directly observed. The method and theory of fMRI are given, one of fMRI data processing software package - statistical parameter mapping (SPM) and its mathematics principle are introduced.The application of ICA to fMRI data is discussed. Not only the areas of the brain activated by the task separated, but also the areas activated by palpitation, eye movement and head movement are gained. Receiver operating characteristic (ROC) is used to evaluate the spatial resolution of the separation of three algorithms, A compare of ICA and SPM is done by ROC.