Deep fuzzy model for non-linear effective connectivity estimation in the intuition of consciousness correlates

Abstract

The brain connectivity, as a promising technique to explore brain networks during resting-states or cognitive tasks, has been employed remarkably in recent years. The aim of this study is to propose a new approach to improve the Granger causality as one of the fundamental methods for calculation of brain effective connectivity. To this end, we utilized a deep fuzzy structure to model the multivariate autoregressive used in the Granger causality. The proposed model benefits from the hierarchical stacked structure where first – order TSK fuzzy rules are the cores of the network. In the first layer of the stacked structure, the antecedents of the fuzzy rules are extracted from the fuzzy clustering of the input space. For subsequent layers, due to the input perturbation which is caused by the previous layer output, a shuffling approach is adopted. To assess our proposed model, we applied it to two nonlinear synthetic time series and compared it with linear Granger. Results revealed that our model is superior in the detection of effective connectivity. We additionally exploit the pioneer model for one of the controversial concepts in cognitive neuroscience in recent years: the neural correlates of visual consciousness. We applied our method to detect connectivity networks of EEG in consciousness states. Our results demonstrated that the proposed nonlinear connectivity estimator was capable of detecting novel correlates: significant differences have been observed among different states of consciousness, not only in presence of attention, as the linear method detected it, but also in absence of attention.