A global coupling index of multivariate neural series with application to the evaluation of mild cognitive impairment

• The GCI method proposed in this paper could be applied to indicate genuine and stochastic synchronization in multivariate EEG series at different frequency bands.
• The GCI method proposed in this paper was less influenced by the frequency bands than the GSI and SS-estimator methods, indicating that the GCI method was more robust. And the GCI method had a better performance on the coupling coefficient relative to GSI and SS-estimator.
• The GCI method proposed in this paper was more sensitive than GSI and SS-estimator methods in differing synchronization strength of EEG between MCI and NC, and could be considered as a potential indicator diagnosing MCI.

Recently, the synchronization between neural signals has been widely used as a key indicator of brain function. To understand comprehensively the effect of synchronization on the brain function, accurate computation of the synchronization strength among multivariate neural series from the whole brain is necessary. In this study, we proposed a method named global coupling index (GCI) to estimate the synchronization strength of multiple neural signals. First of all, performance of the GCI method was evaluated by analyzing simulated EEG signals from a multi-channel neural mass model, including the effects of the frequency band, the coupling coefficient, and the signal noise ratio. Then, the GCI method was applied to analyze the EEG signals from 12 mild cognitive impairment (MCI) subjects and 12 normal controls (NC). The results showed that GCI method had two major advantages over the global synchronization index (GSI) or SS-estimator. Firstly, simulation data showed that the GCI method provided both a more robust result on the frequency band and a better performance on the coupling coefficients. Secondly, the actual EEG data demonstrated that GCI method was more sensitive in differentiating the MCI from control subjects, in terms of the global synchronization strength of neural series of specific alpha, beta1 and beta2 frequency bands. Hence, it is suggested that GCI is a better method over GSI and SS-estimator to estimate the synchronization strength of multivariate neural series for predicting the MCI from the whole brain EEG recordings.