Brain functional network segregation and integration in patients with disorders of consciousness

Introduction: The brain regulates information flow by balancing integration and segregation of networks to facilitate flexible cognition and behavior. However, it is unclear how this mechanism manifests during loss of consciousness [1-3]. In this study, we studied brain network segregation and integration using resting state functional magnetic resonance imaging (fMRI) data to assess brain networks in patients with disorders of consciousness.
Methods: Fifty-four patients with disorders of consciousness (24 unresponsive wakefulness syndrome (UWS) (M:F=16:8; mean age= 45±13), 30 minimally conscious state (MCS) (M:F=23:7; mean age= 36±14) and 30 age- and gender-matched healthy controls underwent fMRI. The resting-state MRI data were acquired during wakefulness with eyes closed using a 3 Tesla MRI scanner. Additionally a T1-weighted, structural imaging was performed for anatomical coregistration. First, the fMRI data were pre-processed for realignment, co-registration, segmentation, normalization, head motion regressed out and 0.01-0.1Hz band pass filtered. Data were then parcellated in 256 brain regions (ROIs) using Shen functional atlas from [4]. The connectivity matrix was computed using Pearson correlation. Graph theory connectivity was carried out to measure brain network topological properties in terms of network segregation and integration by computing binarized undirected connectivity matrix. Normalized clustering coefficients were computed as measures of network segregation while normalized participation coefficients were computed as measures of network integration [3]. Through integrated nodal graph measures, individual networks (such as default mode, frontoparietal, auditory, salience, subcortical and cerebellum networks) were also computed to study which networks were predominantly affected [3]. To enable comparison of network properties across groups, we used sparsity-based threshold (S) to avoid spurious results. To prevent biases associated with a single threshold, we determined a range of sparsity (0.06 ≤ S ≤ 0.5, with an increment of 0.025), which avoids excess network fragmentation at sparser thresholds. The between group differences for global (i.e., whole brain) and individual networks were computed with unpaired t-test with FDR correction for multiple comparison [5-6]. Finally the network segregation and integration mean values were correlated with Coma Recovery Scale-Revised (CRS-R) modified score [7].
Results: Patients in UWS had decreased participation coefficients (network integration) compared to those in MCS (effect size= -0.44, p<0.0001) and controls (effect size= -0.63, p<0.0001). Patients in MCS had significant decreased participation coefficients compared to controls (effect size= -0.37, p<0.001). On the other hand, patients in UWS had significant increased clustering coefficient (network segregation) compared to those in MCS (effect size= 0.39, p= <0.001) and controls (effect size= 0.63, p<0.0001). Patients in MCS had significant increased clustering coefficients compared to controls (effect size= 0.03, p<0.01). This decreased participation coefficient and increased clustering coefficient were noted predominantly observed in the frontoparietal and subcortical networks.
Conclusions: Patients with disorders of consciousness present decreased in network integration and increased in network segregation. Notably, fragmentation of network integration is observed in patients in unaware patients (UWS), which indicates impaired information flow in the brain modules, especially in the frontoparietal and subcortical networks. This introduces a potential measure to classify patients with disorders of consciousness, which could ultimately be used for clinical diagnosis.
Reference:
1. Fukushima, M., (2018). Structure–function relationships during segregated and integrated network states of human brain functional connectivity. Brain Structure and Function, 223(3), 1091-1106. 2. Deco, G., (2015). Rethinking segregation and integration: contributions of whole-brain modelling. Nature Reviews Neuroscience, 16(7), 430. 3. Keerativittayayut, R., (2018). Large-scale network integration in the human brain tracks temporal uctuations in memory encoding performance. eLife, 7, e32696. 4. Finn, E. S., (2015). Functional connectome ngerprinting: identifying individuals using patterns of brain connectivity. Nature neuroscience, 18(11), 1664. 5. Holla, B., (2017). Disrupted resting brain graph measures in individuals at high risk for alcoholism. Psychiatry Research: Neuroimaging, 265, 54-64. 6. Chennu, S., (2017). Brain networks predict metabolism, diagnosis and prognosis at the bedside in disorders of consciousness. Brain, 140(8), 2120-2132. 7. Demertzi, A., (2015). Intrinsic functional connectivity differentiates minimally conscious from unresponsive patients. Brain, 138(9), 2619-2631.