Time-varying EEG networks of major depressive disorder during facial emotion tasks

Abstract

Depression is a mental disease involved in emotional and cognitive impairments. Neuroimaging studies have found abnormalities in the structure and functional network of brain for major depressive disorder (MDD).However, neural mechanism of the dynamic connectivity for emotional attention of MDD is currently insufficient. In this study, event-related potentials (ERP) and time-varying network were analyzed to investigate attention bias and corresponding neural mechanisms induced by emotional facial stimuli. In the ERP results, N100 components in MDD had shorter latencies and smaller amplitudes than those in healthy controls (HC) for sad and fear faces. The P200 amplitudes induced by sad faces in MDD were significantly higher than those induced by happy and fear faces in MDD, and those induced by sad faces in HC. It was indicated that MDD patients had attention bias towards sad faces. For the time-varying network analysis, adaptive directed transfer function was explored to construct dynamic network connectivity. MDD patients had stronger information outflow from the right frontal region and weaker information outflow from parieto-occipital regions for sad faces. In addition, the network properties of sad faces were significantly correlated with PHQ-9 scores for MDD group. These findings may provide further explanation for understanding the MDD's neural mechanism of attention bias during facial emotional tasks.

Keywords: Adaptive directed transfer function (ADTF); Attention bias; Event-related potentials (ERP); Major depressive disorder (MDD); Network connectivity.

Fig. 1

Grand-averaged ERPs of three emotional faces for MDD and HC groups; A ERPs at the Fz, FCz and Cz electrodes; B The scalp topographies of N100 and P200 for MDD and HC groups among three emotional faces

Fig. 2

The amplitudes and latencies of N100 and P200 components among three emotional faces for MDD and HC groups. Black stars stand for significant differences between MDD and HC groups. Blue stars stand for significant differences among three emotional faces for MDD group. Yellow stars stand for significant differences among three emotional faces for HC group. Each blue point stands for the result of each MDD. Each yellow point stands for the result of each HC. * denotes p < 0.05, ** denotes p < 0.01, and *** denotes p < 0.001

Fig. 3

Time-varying networks of sad faces for MDD and HC groups; a network topologies of MDD and HC group, b the edges with significantly strong connection weights in the network topologies. The electrodes with the same color as the lines represent the starting points of information outflow, otherwise the electrodes with the different color from the lines denote the ending points of information outflow

Fig. 4

the network properties of sad face for MDD and HC groups aginst time at different stages; a clustering coefficient; b local efficiency; c global efficiency; d characteristic path length. The red circles with error bars were the mean and standard deviation of network properties for MDD. The purple circles with error bars were the mean and standard deviation of network properties for HC. Black stars stand for significant differences between MDD and HC groups. *denotes p < 0.05, **denotes p < 0.01, and ***denotes p < 0.001.

Fig. 5

Time-varying networks of sad and happy faces for MDD group; a network topologies of sad and happy faces, b the edges with significantly strong connection weights in the network topologies. The electrodes with the same color as the lines represent the starting points of information outflow, otherwise the electrodes with the different color from the lines denote the ending points of information outflow

Fig. 6

The network properties of sad and happy faces for MDD group aginst time at different stages; a clustering coefficient; b local efficiency; c global efficiency; d characteristic path length. The red circles with error bars were the mean and standard deviation of network properties for sad faces. The purple circles with error bars were the mean and standard deviation of network properties for happy faces. Black stars stand for significant differences between sad and happy faces. *denotes p < 0.05, **denotes p < 0.01, and ***denotes p < 0.001

Fig. 7

Pearson's correlation coefficients between network properties and PHQ-9 scores for sad face of MDD group. a pre-attentive stage; b mid-attentive stage; c after-attentive stage. r stands for the correlation coefficient. p stands for the significant level