Abstract:Recent studies show that the dynamics of electrophysiological functional connectivity is attracting more and more interest since it is considered as a better representation of functional brain networks than static network analysis. It is believed that the dynamic electrophysiological brain networks with specific frequency modes, transiently form and dissolve to support ongoing cognitive function during continuous task performance. Here, we propose a novel method based on tensor component analysis (TCA), to characterize the spatial, temporal, and spectral signatures of dynamic electrophysiological brain networks in electroencephalography (EEG) data recorded during free music-listening. A three-way tensor containing time-frequency phase-coupling between pairs of parcellated brain regions is constructed. Nonnegative CANDECOMP/PARAFAC (CP) decomposition is then applied to extract three interconnected, low-dimensional descriptions of data including temporal, spectral, and spatial connection factors. Musical features are also extracted from stimuli using acoustic feature extraction. Correlation analysis is then conducted between temporal courses of musical features and TCA components to examine the modulation of brain patterns. We derive several brain networks with distinct spectral modes (described by TCA components) significantly modulated by musical features, including higher-order cognitive, sensorimotor, and auditory networks. The results demonstrate that brain networks during music listening in EEG are well characterized by TCA components, with spatial patterns of oscillatory phase-synchronization in specific spectral modes. The proposed method provides evidence for the time-frequency dynamics of brain networks during free music listening through TCA, which allows us to better understand the reorganization of electrophysiological networks.

Measuring the Task Induced Oscillatory Brain Activity Using Tensor Decomposition.

Detecting Brain Activations by Constrained Non-Negative Matrix Factorization from Task-Related BOLD Fmri

Discovering Dynamic Task-Modulated Functional Networks with Specific Spectral Modes Using MEG

Tensor Decomposition of EEG Signals: A Brief Review

Tensor Decomposition for EEG Signal Retrieval

Multi-domain Features of the Non-phase-locked Component of Interest Extracted from ERP Data by Tensor Decomposition

Subject-Specific Modeling of EEG-fNIRS Neurovascular Coupling by Task-Related Tensor Decomposition

A Comparison of Non-negative Tucker Decomposition and Parallel Factor Analysis for Identification and Measurement of Human EEG Rhythms

Tensor Decomposition of Large-scale Clinical EEGs Reveals Interpretable Patterns of Brain Physiology

Analysis of Ongoing EEG Elicited by Natural Music Stimuli Using Nonnegative Tensor Factorization

Sparse Nonnegative Tensor Decomposition for EEG Data

Exploring the feasibility of tensor decomposition for analysis of fNIRS signals: a comparative study with grand averaging method

Discovering hidden brain network responses to naturalistic stimuli via tensor component analysis of multi-subject fMRI data

Dynamic Network Analysis of Electrophysiological Task Data

Slice Oriented Tensor Decomposition of EEG Data for Feature Extraction in Space, Frequency and Time Domains

Extracting Multi-Mode ERP Features Using Fifth-Order Nonnegative Tensor Decomposition

Increasing Stability of EEG Components Extraction Using Sparsity Regularized Tensor Decomposition.

Using an ODE model to separate Rest and Task signals in fMRI

Deriving Electrophysiological Brain Network Connectivity Via Tensor Component Analysis During Freely Listening to Music

Task-Independent EEG Identification Via Low-Rank Matrix Decomposition

Sequential nonnegative tucker decomposition on multi-way array of time-frequency transformed event-related potentials