Disclosing large-scale directed functional connections in MEG with the multivariate phase slope index

Abstract

The phase slope index (PSI) is a method to disclose the direction of frequency-specific neural interactions from magnetoencephalographic (MEG) time series. A fundamental property of PSI is that of vanishing for linear mixing of independent neural sources. This property allows PSI to cope with the artificial instantaneous connectivity among MEG sensors or brain sources induced by the field spread. Nevertheless, PSI is limited by being a bivariate estimator of directionality as opposite to the multidimensional nature of brain activity as revealed by MEG. The purpose of this work is to provide a multivariate generalization of PSI. We termed this measure as the multivariate phase slope index (MPSI). In order to test the ability of MPSI in estimating the directionality, and to compare the MPSI results to those obtained by bivariate PSI approaches based on maximizing imaginary part of coherency and on canonical correlation analysis, we used extensive simulations. We proved that MPSI achieves the highest performance and that in a large number of simulated cases, the bivariate methods, as opposed to MPSI, do not detect a statistically significant directionality. Finally, we applied MPSI to assess seed-based directed functional connectivity in the alpha band from resting state MEG data of 61 subjects from the Human Connectome Project. The obtained results highlight a directed functional coupling in the alpha band between the primary visual cortex and several key regions of well-known resting state networks, e.g. dorsal attention network and fronto-parietal network.

Brain Functional Connectivity Through Phase Coupling of Neuronal Oscillations: A Perspective From Magnetoencephalography

“…To assess the directionality from pairs of vector-signals, Basti et al (2018) generalized the definition of PSI to multivariate time series, called multivariate phase slope index (MPSI). MPSI is defined as…”

“…MPSI solely detects the directionality of phase-lagged coupling; a positive value of MPSI indicates that the vector-signal I leads the vector-signal J , while a negative value indicates the opposite. Moreover, similarly to MIM, MPSI is invariant under invertible and static linear transformations, and thus it is independent on rotations of the physical coordinate system of MEG source space (Basti et al, 2018). In the case of two univariate time series, MPSI coincides with PSI, apart from a normalization factor.…”

“…More recently, our group introduced a novel phase coherence metric, namely the Multivariate Phase Slope Index, to investigate the directionality of functional coupling in MEG and EEG data (Basti et al, 2018). This novel metric has been used to investigate the directionality of functional coupling in the alpha frequency band between the visual network and the whole brain in resting state MEG data from a large cohort of healthy subjects from the Human Connectome Project (Larson-Prior et al, 2013).…”

Analysing linear multivariate pattern transformations in neuroimaging data

“…by averaging across voxels or by selecting the directions which explain the highest variance (PCA). This process leads to a loss of information and potentially to biased connectivity estimates (Marzetti et al 2013;Geerligs et al 2016;Anzellotti et al 2017Anzellotti et al , 2018Basti et al 2018). Importantly, it also makes it impossible to estimate the transformations between patterns among different ROIs, and to describe functionally relevant features of those mappings.…”

Simultaneous MEG/EEG recordings for the study of source domain brain connectivity in neurodegenerative diseases

“…This matrix has dimension , and contains entries for all the bivariate synchronizations between all passible pairs of time series, one from each set. This method is, for example, used to extend the Phase Slope Index (Nolte et al, 2008) into its multivariate counterpart (Basti et al, 2018).…”