Networked information interactions in schizophrenia magnetoencephalograms based on permutation transfer entropy

Brain network science plays an important role in the exploration of neurological and psychiatric diseases. To explore the information interactions in the magnetoencephalogram (MEG) data of schizophrenia, we construct resting-state brain networks based on permutation transfer entropy in 17 schizophrenia patients (SCZs) and 14 healthy controls (HCs). In this process, the effects of equal values on permutation and probability distribution are particularly considered. We quantify three network features, namely, weight, complexity and nonequilibrium, to characterize the schizophrenia MEG network. The results indicate that the level of information interactions between brain regions, the inward, outward and total information flows in brain regions of SCZs are generally smaller than those of HCs. In the complexity analysis, the Shannon entropy of the information inflows and outflows in the right parietal region of SCZs exhibits significant differences, and further the Shannon entropy of the information inflows (p=0.003) and outflows (p=0.034) of the whole brain of SCZs is significantly higher than that for HCs. Additionally, SCZs have lower values of local nonequilibrium than HCs in all brain regions except the middle central and middle frontal regions, with the right parietal region having the strongest significance (p=0.015), and the whole-brain nonequilibrium of SCZs is significantly lower than that of HCs (p=0.022). The MEG network constructed based on permutation transfer entropy can be used to effectively extract the characteristics of schizophrenia, and the comparative analysis of the complexity and nonequilibrium features can expand the exploration of the pathological and physiological mechanisms of schizophrenia.