Controlling Dimensionality in a Systems Approach to Dynamic Multimodal Functional Brain Imaging

The complementary spatial, temporal and specificity advantages of fMRI, EEG, MEG, PET and DOT for functional brain imaging motivate interest in multimodal functional brain imaging. State-variable dynamical systems modeling of neural activity and its relation to local hemodynamics, further coupled with autonomic physiology offers enhanced spatiotemporal resolution and insight into physiological signals and mechanisms. However, such a model also implies an explosion of state dimension. We discuss strategies for controlling this high dimensionality based on subspace approaches applied to the observed data and the model structure.