| Modern neuromagnetic source imaging techniques provide a unique opportunity to map the electrical activity of the human brain non-invasively with a high temporal and spatial resolution. Using these methods, we can begin to understand how brain systems work together as a differentiated yet integrated dynamical whole. The main hypothesis in this thesis is that functionally coupled ensembles of neurons are spontaneously active due to their intrinsic cellular and network properties, and that sensory inputs modulate their synchronization and coherence at multiple frequency bands. Computational methods for current density reconstruction, blind source separation, time-frequency analysis, and data visualization are used to characterize the spontaneous and evoked dynamics of the human brain.; Anatomically constrained recursive weighted minimum-norm inverse algorithms are developed that find noise-regularized sparse solutions by minimizing diversity measures. Computer simulations are used to determine which initialization functions and diversity measures are optimal as quantified by several reconstruction and spatial error metrics. The effects of noise and the benefits of using a spatiotemporal extension of the algorithm that simultaneously solves multiple measurement vectors are also characterized. A new recursive adaptive minimum-norm algorithm is developed that does not require an a priori weight matrix but rather adaptively optimizes the recursive weights based on the resolution matrix. Simulations show that this algorithm represents a major improvement over the non-adaptive algorithms. Independent Component Analysis is demonstrated to dramatically increase the number of recovered sources. We focus on three time-frequency measures: (1) the Event-Related Spectral Perturbation (ERSP), (2) the Inter-Trial Phase Coherence (ITPC), and (3) the Inter-Trial Phase Cross Coherence (ITPCC).; With these methods, we characterize the spontaneously active independent sources of the brain, and investigate how they are modulated by visual, somatosensory, auditory, multimodal, and apparent motion stimulation. We characterize the patterns of ERSP, ITPC, and ITPCC at the theta, alpha, beta, and gamma bands. Applications in clinical neuroscience are demonstrated in patients suffering from psychosis, depression, obsessive-compulsive disorder, pain, and Parkinson's disease. |
