Fig 1.
Neural correlates of salience processing defined with the EEG single-trial variability (STV) informed fMRI analysis.
(A) Timing diagram showing significant group-level activation clusters (p < 0.05 cluster-wise multiple comparison correction). STV in EEG temporal components discriminating the target versus standard trials was used to map the spatiotemporally distributed BOLD fMRI correlates spanning the trial. EEG STV information was incorporated as BOLD predictors in voxel-wise general linear model (GLM) analysis of fMRI, controlling for the variance due to the presence of stimuli and response time (RT). Cluster colors denote positive (red) and negative (blue) effects. Time is relative to stimulus onset. (B) Definition of salience processing nodes. Each node is a sphere centered on the peak voxel of the group-level STV EEG-informed fMRI analysis results. Centroid of peak locations was used for regions involved in more than one temporal windows. Node colors denote timing of involvement in the trial from early to late (temporal order: red, orange, yellow, green, and blue). All clusters and nodes were overlaid on a 3D Montreal Neurological Institute (MNI) 152 brain pial surface for visualization. BOLD, blood-oxygen-level-dependent; RH, right hemisphere; LH, left hemisphere; A, anterior; P, posterior; S, superior; I inferior; SPL, superior parietal lobule; M1, primary motor cortex; S1, primary somatosensory cortex; V2, secondary visual cortex; OFC, orbitofrontal cortex; IPL, inferior parietal lobule; IFC, inferior frontal cortex; mPFC, medial prefrontal cortex; SMA, supplementary motor area.
Fig 2.
Network localization approach to map functional networks underlying salience processing nodes.
(A) BOLD signals from the nodes (intersected with the gray matter mask) were extracted, controlling the nuisance signals (motion-related, ventricle, and white matter signals). (B) Group-level functional connectivity (FC) results of each node (t-value, mixed-effect, p < 0.001 uncorrected). Seed-based FC analysis (with the task-related variability regressed out) was used to map network of regions connected to each node location. Colors denote positive (red) and negative (blue) correlations. (C) Spatial overlaps in the FC maps of each node identified spatial network organizations of salience processing nodes. Colors represent the number of FC maps overlapped. lSPL and rSPL, left and right superior parietal lobule; rM1, right primary motor cortex; lS1, left primary somatosensory cortex; rV2, right secondary visual area; lOFC and rOFC, left and right orbitofrontal cortex; lIPL, left inferior parietal lobule; rIFC, right inferior frontal cortex.
Fig 3.
Functional connectivity (FC) across salience processing nodes (group averaged z-score, mixed-effect, p < 0.05 uncorrected).
fMRI BOLD signals from the nodes (intersected with the gray matter mask) were extracted, controlling the nuisance signals (motion-related, ventricle and white matter signals). Pearson’s correlation was calculated between BOLD signals from the nodes (with the task-related variability regressed out). FC results identified three distinct groups of the nodes, organized by the EEG discriminating component time windows, indicating a temporal network organization of the nodes: 1) early-time network includes lSPL and rSPL, rM1, rV2, and lS1; 2) middle-time network includes lOFC and rOFC, lIPL, and rIFC; 3) late-time network includes mPFC, SMA, left frontal operculum and temporal pole.
Fig 4.
Effective connectivity (EC) across salience processing nodes (Bayesian parameter averaging, α < 0.05, Bonferroni corrected).
(A) positive EC, (B) negative EC. By leveraging the high temporal information in the EEG data, an effective connectivity state-space model was fit with the salience processing nodes. The arrow and thickness of the connecting lines correspond to the directionality and strength of EC, respectively. Dominant influence is observed in the connections of lSPL, lOFC, and mPFC-SMA. The results here reflect mean group effect. Node colors denote timing of involvement (early-time: red; middle-time: green; late-time: blue).
Fig 5.
Brain-pupil relationships of the cortical network-level effective connectivity (EC) and task-evoked pupillary response (TEPR) in salience processing.
(A) The oddball-modulated positive EC strength from the late-time to early-time network correlated with higher TEPR of oddball trials (p < 0.0035). In (B) and (C), to test the associations between pupil measurements and the triple-network model (SN, salience network; DAN, dorsal attention network; DMN, default mode network), we computed EC across nodes of these networks. (B) The oddball-modulated positive EC strength from SN to DAN correlated with higher TEPR of oddball trials (p < 0.0013). (C) The oddball-modulated negative EC strength from SN to DMN correlated with higher TEPR of oddball trials (p < 0.0060).
Fig 6.
Neural cascades of salience processing and the spatiotemporal network organizations of salience processing nodes.
Previous seed-based and node-by-node functional connectivity results suggest both spatial and temporal network organizations of the identified salience processing nodes, respectively. We hypothesized that the node-specific involvement of these functional networks might indicate a crucial role of these nodes in the temporally evolved processes of salience signal and the relationships between these networks. ECN, executive control network.
Fig 7.
Cortico-subcortical integrated network reorganization (CS-INR) system.
Previous brain-pupil relationships results aligned with the network switching model of SN in the literature, and also showed the role of the locus coeruleus norepinephrine (LC-NE) system in the network reset and the dynamic switching between anticorrelated networks (SN-to-DAN and SN-to-DMN). In support with the literature [14], we hypothesized that the reset and switching might be modulated by the release of the NE, as an effect of the ascending neuromodulation, which indicates that the SN and LC-NE system might cooperate and share an integrated role in salience processing. dACC, dorsal anterior cingulate cortex; lAI and rAI, left and right anterior insula; lSPL and rSPL, left and right superior parietal lobule; lFEF and rFEF, left and right frontal eye fields; mPFC, medial prefrontal cortex; PCC, posterior cingulate cortex; lAG and rAG, left and right angular gyrus.