Dynamic segregation and integration of brain functional networks associated with emotional arousal

Figures

• 

  Figure 1. Analysis schematic illustration EEG data were filtered into four frequency bands. (A) The sliding window method was used to construct dFC by calculating the PLV between channels. (B) For each FC network, the nested-spectral partition (NSP) method was applied to obtain the hierarchical modules that correspond to eigenvectors with different orders (here shown the first three levels), and the integration component Hin, segregation component Hse and their competition HB = Hin-Hse were obtained. Based on the time series of dynamic Hin, Hse, and HB, we defined the dynamic brain network measures integration strength Min, segregation strength Mse. Dwell time at integration state Tin and switch frequency between segregation and integration states fIS. These measures were mapped to regions and thus the regional analysis can be performed.

• 

  Figure 2. Normalized distributions of network measures on regions during emotional stimuli, averaged across all subjects and trials

• 

  Figure 3. Time-averaged integration and segregation at different arousal states (A) Integration strength Min (upper) and segregation strength Mse (lower) under two arousal conditions. ∗p < 0.05. (B) The difference in regional Min (upper) and Mse (lower) between HA and LA conditions in four frequency bands. The regions with significant alterations were highlighted with yellow points (p < 0.05, FDR corrected).

• 

  Figure 4. Dynamic analysis of the brain under HA and LA conditions (A) Dwell time at integration states (upper) and switching frequency (lower). ∗∗p < 0.01. (B) Regional differences in dwell time and switching frequency between the HA and LA conditions. The electrodes with significant changes are highlighted with yellow points (p < 0.05, FDR corrected).

• 

  Figure 5. Correlations between brain network measures and arousal ratings The highlighted electrodes have significant correlations (p < 0.05, FDR corrected).

• 

  Figure 6. Weights of regions in the prediction models All of the features were normalized, and the positive/negative weights contributed to higher/lower individual arousal variability.