Cross-Predictions in the Search for Effective Connectivity in Brain

Complex networks are ubiquitous in the real world. Brain activity, represented by multi-channel electroencephalographic (EEG) signals, is one of the most topical examples. Modern methods are beginning to successfully detect effective connectivity (meaning causal interactions, not structural or functional relations) in the brain. In this study, cross-predictions in reconstructed state spaces are used for bivariate causal detection and also for exploring dynamical networks whose nodes are characterized by time series. The method is applied to EEG signals recorded from six positions on the head during an experiment with visual stimulation (VS) of the brain. An intense causal influence from the back to the frontal parts of the same hemisphere, most pronounced in the occipital areas of the cortex, is detected. The observed causal effect persists for a short time even after VS is switched off.