A Method for Investigating the Nonlinear Dynamics of the Human Brain from Analysis of Functional Mri Data

Working with massive fMRI data is difficult because of the poor signal to noise ratio in the data, lack of a statistical distribution for the uncontaminated data, and not having a robust statistical model for the noise. Bayesian methods have the advantage that they make our simplifying assumptions explicit in the model. Nonetheless, the Bayesian models will reach their limits in the absence of better estimates on the distribution of data or at least, partial knowledge of such distributions. We propose a method for obtaining relevant probabilistic distributions by extracting information about nonlinearity in those representations of data that lend themselves to certain operations that we call for short "local-to-global integration" and "local linearization". We apply this technique to fMRI data to describe certain aspects in nonlinear dynamics of the brain function as reported by the fMRI acquisition