Estimating the in-mean and in-variance connectomes of the human brain.
Duggento A., Passamonti L., Guerrisi M., Toschi N.
Fluctuations in the fMRI signal contain important information about neurovascular coupling and connectivity. We develop a framework for estimating both in-mean and in-variance causality in complex networks through exponentially weighted moving average/variance techniques followed by conditioned Granger causality. For validation, we evolve synthetic networks of Kuramoto oscillators with dynamical GARCH noise. In human data, we provide the first estimate of the in-variance human connectome, revealing the existence of information exchange in the opposite direction with respect to traditional connectivity estimates. Our methods provide access to a novel brain connectivity layer which complements the current knowledge in a physiologically meaningful manner.