Higher-order functional connectivity analysis of resting-state functional magnetic resonance imaging data using multivariate cumulants.

Blood-level oxygenation-dependent (BOLD) functional magnetic resonance imaging (fMRI) is the most common modality to study functional connectivity in the human brain. Most research to date has focused on connectivity between pairs of brain regions. However, attention has recently turned towards connectivity involving more than two regions, that is, higher-order connectivity. It is not yet clear how higher-order connectivity can best be quantified. The measures that are currently in use cannot distinguish between pairwise (i.e., second-order) and higher-order connectivity. We show that genuine higher-order connectivity can be quantified by using multivariate cumulants. We explore the use of multivariate cumulants for quantifying higher-order connectivity and the performance of block bootstrapping for statistical inference. In particular, we formulate a generative model for fMRI signals exhibiting higher-order connectivity and use it to assess bias, standard errors, and detection probabilities. Application to resting-state fMRI data from the Human Connectome Project demonstrates that spontaneous fMRI signals are organized into higher-order networks that are distinct from second-order resting-state networks. Application to a clinical cohort of patients with multiple sclerosis further demonstrates that cumulants can be used to classify disease groups and explain behavioral variability. Hence, we present a novel framework to reliably estimate genuine higher-order connectivity in fMRI data which can be used for constructing hyperedges, and finally, which can readily be applied to fMRI data from populations with neuropsychiatric disease or cognitive neuroscientific experiments.

Network Analysis in Multiple Sclerosis and Related Disorders.

Multiple sclerosis (MS) is a neuroinflammatory and neurodegenerative disease of the central nervous system, commonly featuring disability and cognitive impairment. The pathologic hallmark of MS lies in demyelination and hence impaired structural and functional neuronal pathways. Recent studies have shown that MS shows extensive structural disconnection of key network hub areas like the thalamus, combined with a functional network reorganization that can mostly be related to poorer clinical functioning. As MS can, therefore, be considered a network disorder, this review outlines recent innovations in the field of network neuroscience in MS.

Eigenvector Centrality Mapping Reveals Volatility of Functional Brain Dynamics in Anti-NMDA Receptor Encephalitis.

BACKGROUND: Anti-NMDA receptor encephalitis (NMDARE) causes long-lasting cognitive deficits associated with altered functional connectivity. Eigenvector centrality (EC) mapping represents a powerful new method for data-driven voxelwise and time-resolved estimation of network importance-beyond changes in classical static functional connectivity. METHODS: To assess changes in functional brain network organization, we applied EC mapping in 73 patients with NMDARE and 73 matched healthy control participants. Areas with significant group differences were further investigated using 1) spatial clustering analyses, 2) time series correlation to assess synchronicity between the hippocampus and cortical brain regions, and 3) correlation with cognitive and clinical parameters. RESULTS: Dynamic, time-resolved EC showed significantly higher variability in 13 cortical areas (familywise error p < .05) in patients with NMDARE compared with healthy control participants. Areas with dynamic EC group differences were spatially organized in centrality clusters resembling resting-state networks. Importantly, variability of dynamic EC in the frontotemporal cluster was associated with impaired verbal episodic memory in patients (r = -0.25, p = .037). EC synchronicity between the hippocampus and the medial prefrontal cortex was reduced in patients compared with healthy control participants (familywise error p < .05, tmax = 3.76) and associated with verbal episodic memory in patients (r = 0.28, p = .019). Static EC analyses showed group differences in only one brain region (left intracalcarine cortex). CONCLUSIONS: Widespread changes in network dynamics and reduced hippocampal-medial prefrontal synchronicity were associated with verbal episodic memory deficits and may thus represent a functional neural correlate of cognitive dysfunction in NMDARE. Importantly, dynamic EC detected substantially more network alterations than traditional static approaches, highlighting the potential of this method to explain long-term deficits in NMDARE.

The cognitive relevance of non-lesional damage to cortical networks in people with multiple sclerosis.

BACKGROUND: Cognitive impairment, a common and debilitating symptom in people with multiple sclerosis (MS), is especially related to cortical damage. However, the impact of regional cortical damage remains poorly understood. Our aim was to evaluate structural (network) integrity in lesional and non-lesional cortex in people with MS, and its relationship with cognitive dysfunction. METHODS: In this cross-sectional study, 176 people with MS and 48 healthy controls underwent MRI, including double inversion recovery and diffusion-weighted scans, and neuropsychological assessment. Cortical integrity was assessed based on fractional anisotropy (FA) and mean diffusivity (MD) within 212 regions split into lesional or non-lesional cortex, and grouped into seven cortical networks. Integrity was compared between people with MS and controls, and across cognitive groups: cognitively-impaired (CI; ≥ two domains at Z ≤ - 2 below controls), mildly CI (≥ two at - 2 < Z ≤ - 1.5), or cognitively-preserved (CP). RESULTS: Cortical lesions were observed in 87.5% of people with MS, mainly in ventral attention network, followed by limbic and default mode networks. Compared to controls, in non-lesional cortex, MD was increased in people with MS, but mean FA did not differ. Within the same individual, MD and FA were increased in lesional compared to non-lesional cortex. CI-MS exhibited higher MD than CP-MS in non-lesional cortex of default mode, frontoparietal and sensorimotor networks, of which the default mode network could best explain cognitive performance. CONCLUSION: Diffusion differences in lesional cortex were more severe than in non-lesional cortex. However, while most people with MS had cortical lesions, diffusion differences in CI-MS were more prominent in non-lesional cortex than lesional cortex, especially within default mode, frontoparietal and sensorimotor networks.

Energy Associated With Dynamic Network Changes in Patients With Multiple Sclerosis and Cognitive Impairment.

BACKGROUND AND OBJECTIVES: Patients with multiple sclerosis (MS) often experience cognitive impairment, and this is related to structural disconnection and subsequent functional reorganization. It is unclear how specific patterns of functional reorganization might make it harder for cognitively impaired (CI) patients with MS to dynamically adapt how brain regions communicate, which is crucial for normal cognition. We aimed to identify dynamic functional network patterns that are relevant to cognitive impairment in MS and investigate whether these patterns can be explained by altered energy costs. METHODS: Resting-state functional and diffusion MRI was acquired in a cross-sectional design, as part of the Amsterdam MS cohort. Patients with clinically definitive MS (relapse-free) were classified as CI (≥2/7 domains Z < -2), mildly CI (MCI) (≥2/7 domains Z < -1.5), or cognitively preserved (CP) based on an expanded Brief Repeatable Battery of Neuropsychological Tests. Functional connectivity states were determined using k-means clustering of moment-to-moment cofluctuations (i.e., edge time series), and the resulting state sequence was used to characterize the frequency of transitions. Control energy of the state transitions was calculated using the structural network with network control theory. RESULTS: Imaging and cognitive data were available for 95 controls and 330 patients (disease duration: 15 years; 179 CP, 65 MCI, and 86 CI). We identified a "visual network state," "sensorimotor network state," "ventral attention network state," and "default mode network state." CI patients transitioned less frequently between connectivity states compared with CP (ß = -5.78; p = 0.038). Relative to the time spent in a state, CI patients transitioned less from a "default mode network state" to a "visual network state" (ß = -0.02; p = 0.004). The CI patients required more control energy to transition between states (ß = 0.32; p = 0.007), particularly for the same transition (ß = 0.34; p = 0.049). DISCUSSION: This study showed that it costs more energy for MS patients with cognitive impairment to dynamically change the functional network, possibly explaining why these transitions occur less frequently. In particular, transitions from a default mode network state to a visual network state were relevant for cognition in these patients. To further study the order of events leading to these network disturbances, future work should include longitudinal data across different disease stages.

Pathologic Substrates of Structural Brain Network Resilience and Topology in Parkinson Disease Decedents.

BACKGROUND AND OBJECTIVES: In Parkinson disease (PD), α-synuclein spreading through connected brain regions leads to neuronal loss and brain network disruptions. With diffusion-weighted imaging (DWI), it is possible to capture conventional measures of brain network organization and more advanced measures of brain network resilience. We aimed to investigate which neuropathologic processes contribute to regional network topologic changes and brain network resilience in PD. METHODS: Using a combined postmortem MRI and histopathology approach, PD and control brain donors with available postmortem in situ 3D T1-weighted MRI, DWI, and brain tissue were selected from the Netherlands Brain Bank and Normal Aging Brain Collection Amsterdam. Probabilistic tractography was performed, and conventional network topologic measures of regional eigenvector centrality and clustering coefficient, and brain network resilience (change in global efficiency upon regional node failure) were calculated. PSer129 α-synuclein, phosphorylated-tau, ß-amyloid, neurofilament light-chain immunoreactivity, and synaptophysin density were quantified in 8 cortical regions. Group differences and correlations were assessed with rank-based nonparametric tests, with age, sex, and postmortem delay as covariates. RESULTS: Nineteen clinically defined and pathology-confirmed PD (7 F/12 M, 81 ± 7 years) and 15 control (8 F/7 M, 73 ± 9 years) donors were included. With regional conventional measures, we found lower eigenvector centrality only in the parahippocampal gyrus in PD (d = -1.08, 95% CI 0.003-0.010, p = 0.021), which did not associate with underlying pathology. No differences were found in regional clustering coefficient. With the more advanced measure of brain network resilience, we found that the PD brain network was less resilient to node failure of the dorsal anterior insula compared with the control brain network (d = -1.00, 95% CI 0.0012-0.0015, p = 0.018). This change was not directly driven by neuropathologic processes within the dorsal anterior insula or in connected regions but was associated with higher Braak α-synuclein staging (rs = -0.40, p = 0.036). DISCUSSION: Although our cohort might suffer from selection bias, our results highlight that regional network disturbances are more complex to interpret than previously believed. Regional neuropathologic processes did not drive regional topologic changes, but a global increase in α-synuclein pathology had a widespread effect on brain network reorganization in PD.

Increased flexibility of brain dynamics in patients with multiple sclerosis.

Patients with multiple sclerosis consistently show widespread changes in functional connectivity. Yet, alterations are heterogeneous across studies, underscoring the complexity of functional reorganization in multiple sclerosis. Here, we aim to provide new insights by applying a time-resolved graph-analytical framework to identify a clinically relevant pattern of dynamic functional connectivity reconfigurations in multiple sclerosis. Resting-state data from 75 patients with multiple sclerosis (N = 75, female:male ratio of 3:2, median age: 42.0 ± 11.0 years, median disease duration: 6 ± 11.4 years) and 75 age- and sex-matched controls (N = 75, female:male ratio of 3:2, median age: 40.2 ± 11.8 years) were analysed using multilayer community detection. Local, resting-state functional system and global levels of dynamic functional connectivity reconfiguration were characterized using graph-theoretical measures including flexibility, promiscuity, cohesion, disjointedness and entropy. Moreover, we quantified hypo- and hyper-flexibility of brain regions and derived the flexibility reorganization index as a summary measure of whole-brain reorganization. Lastly, we explored the relationship between clinical disability and altered functional dynamics. Significant increases in global flexibility (t = 2.38, PFDR = 0.024), promiscuity (t = 1.94, PFDR = 0.038), entropy (t = 2.17, PFDR = 0.027) and cohesion (t = 2.45, PFDR = 0.024) were observed in patients and were driven by pericentral, limbic and subcortical regions. Importantly, these graph metrics were correlated with clinical disability such that greater reconfiguration dynamics tracked greater disability. Moreover, patients demonstrate a systematic shift in flexibility from sensorimotor areas to transmodal areas, with the most pronounced increases located in regions with generally low dynamics in controls. Together, these findings reveal a hyperflexible reorganization of brain activity in multiple sclerosis that clusters in pericentral, subcortical and limbic areas. This functional reorganization was linked to clinical disability, providing new evidence that alterations of multilayer temporal dynamics play a role in the manifestation of multiple sclerosis.

The network collapse in multiple sclerosis: An overview of novel concepts to address disease dynamics.

Multiple sclerosis is a neuroinflammatory and neurodegenerative disorder of the central nervous system that can be considered a network disorder. In MS, lesional pathology continuously disconnects structural pathways in the brain, forming a disconnection syndrome. Complex functional network changes then occur that are poorly understood but closely follow clinical status. Studying these structural and functional network changes has been and remains crucial to further decipher complex symptoms like cognitive impairment and physical disability. Recent insights especially implicate the importance of monitoring network hubs in MS, like the thalamus and default-mode network which seem especially hit hard. Such network insights in MS have led to the hypothesis that as the network continues to become disconnected and dysfunctional, exceeding a certain threshold of network efficiency loss leads to a "network collapse". After this collapse, crucial network hubs become rigid and overloaded, and at the same time a faster neurodegeneration and accelerated clinical (and cognitive) progression can be seen. As network neuroscience has evolved, the MS field can now move towards a clearer classification of the network collapse itself and specific milestone events leading up to it. Such an updated network-focused conceptual framework of MS could directly impact clinical decision making as well as the design of network-tailored rehabilitation strategies. This review therefore provides an overview of recent network concepts that have enhanced our understanding of clinical progression in MS, especially focusing on cognition, as well as new concepts that will likely move the field forward in the near future.

A more unstable resting-state functional network in cognitively declining multiple sclerosis.

Cognitive impairment is common in people with multiple sclerosis and strongly affects their daily functioning. Reports have linked disturbed cognitive functioning in multiple sclerosis to changes in the organization of the functional network. In a healthy brain, communication between brain regions and which network a region belongs to is continuously and dynamically adapted to enable adequate cognitive function. However, this dynamic network adaptation has not been investigated in multiple sclerosis, and longitudinal network data remain particularly rare. Therefore, the aim of this study was to longitudinally identify patterns of dynamic network reconfigurations that are related to the worsening of cognitive decline in multiple sclerosis. Resting-state functional MRI and cognitive scores (expanded Brief Repeatable Battery of Neuropsychological tests) were acquired in 230 patients with multiple sclerosis and 59 matched healthy controls, at baseline (mean disease duration: 15 years) and at 5-year follow-up. A sliding-window approach was used for functional MRI analyses, where brain regions were dynamically assigned to one of seven literature-based subnetworks. Dynamic reconfigurations of subnetworks were characterized using measures of promiscuity (number of subnetworks switched to), flexibility (number of switches), cohesion (mutual switches) and disjointedness (independent switches). Cross-sectional differences between cognitive groups and longitudinal changes were assessed, as well as relations with structural damage and performance on specific cognitive domains. At baseline, 23% of patients were cognitively impaired (≥2/7 domains Z < -2) and 18% were mildly impaired (≥2/7 domains Z < -1.5). Longitudinally, 28% of patients declined over time (0.25 yearly change on ≥2/7 domains based on reliable change index). Cognitively impaired patients displayed more dynamic network reconfigurations across the whole brain compared with cognitively preserved patients and controls, i.e. showing higher promiscuity (P = 0.047), flexibility (P = 0.008) and cohesion (P = 0.008). Over time, cognitively declining patients showed a further increase in cohesion (P = 0.004), which was not seen in stable patients (P = 0.544). More cohesion was related to more severe structural damage (average r = 0.166, P = 0.015) and worse verbal memory (r = -0.156, P = 0.022), information processing speed (r = -0.202, P = 0.003) and working memory (r = -0.163, P = 0.017). Cognitively impaired multiple sclerosis patients exhibited a more unstable network reconfiguration compared to preserved patients, i.e. brain regions switched between subnetworks more often, which was related to structural damage. This shift to more unstable network reconfigurations was also demonstrated longitudinally in patients that showed cognitive decline only. These results indicate the potential relevance of a progressive destabilization of network topology for understanding cognitive decline in multiple sclerosis.

Functional network dynamics and decreased conscientiousness in multiple sclerosis.

BACKGROUND: Conscientiousness is a personality trait that declines in people with multiple sclerosis (PwMS) and its decline predicts worse clinical outcomes. This study aims to investigate the neural underpinnings of lower Conscientiousness in PwMS by examining MRI anomalies in functional network dynamics. METHODS: 70 PwMS and 50 healthy controls underwent personality assessment and resting-state MRI. Associations with dynamic functional network properties (i.e., eigenvector centrality) were evaluated, using a dynamic sliding-window approach. RESULTS: In PwMS, lower Conscientiousness was associated with increased variability of centrality in the left insula (tmax = 4.21) and right inferior parietal lobule (tmax = 3.79); a relationship also observed in regressions accounting for handedness, disease duration, disability, and tract disruption in relevant structural networks (ΔR2 = 0.071, p = 0.003; ΔR2 = 0.094, p = 0.004). Centrality dynamics of the observed regions were not associated with Neuroticism (R2 < 0.001, p = 0.956; R2 < 0.001, p = 0.945). As well, higher Conscientiousness was associated with greater variability in connectivity for the left insula with the default-mode network (F = 3.92, p = 0.023) and limbic network (F = 5.66, p = 0.005). CONCLUSION: Lower Conscientiousness in PwMS was associated with increased variability in network centrality, most prominently for the left insula and right inferior parietal cortex. This effect, specific to Conscientiousness and significant after accounting for disability and structural network damage, could indicate that overall stable network centrality is lost in patients with low Conscientiousness, especially for the insula and right parietal cortex. The positive relationship between Conscientiousness and variability of connectivity between left insula and default-mode network potentially affirms that dynamics between the salience and default-mode networks is related to the regulation of behavior.

Longitudinal Network Changes and Conversion to Cognitive Impairment in Multiple Sclerosis.

OBJECTIVE: To characterize functional network changes related to conversion to cognitive impairment in a large sample of patients with multiple sclerosis (MS) over a period of 5 years. METHODS: Two hundred twenty-seven patients with MS and 59 healthy controls of the Amsterdam MS cohort underwent neuropsychological testing and resting-state fMRI at 2 time points (time interval 4.9 ± 0.9 years). At both baseline and follow-up, patients were categorized as cognitively preserved (CP; n = 123), mildly impaired (MCI; z < -1.5 on ≥2 cognitive tests, n = 32), or impaired (CI; z < -2 on ≥2 tests, n = 72), and longitudinal conversion between groups was determined. Network function was quantified with eigenvector centrality, a measure of regional network importance, which was computed for individual resting-state networks at both time points. RESULTS: Over time, 18.9% of patients converted to a worse phenotype; 22 of 123 patients who were CP (17.9%) converted from CP to MCI, 10 of 123 from CP to CI (8.1%), and 12 of 32 patients with MCI converted to CI (37.5%). At baseline, default-mode network (DMN) centrality was higher in CI individuals compared to controls (p = 0.05). Longitudinally, ventral attention network (VAN) importance increased in CP, driven by stable CP and CP-to-MCI converters (p < 0.05). CONCLUSIONS: Of all patients, 19% worsened in their cognitive status over 5 years. Conversion from intact cognition to impairment is related to an initial disturbed functioning of the VAN, then shifting toward DMN dysfunction in CI. Because the VAN normally relays information to the DMN, these results could indicate that in MS normal processes crucial for maintaining overall network stability are progressively disrupted as patients clinically progress.

Dynamic functional connectivity as a neural correlate of fatigue in multiple sclerosis.

BACKGROUND: More than 80% of multiple sclerosis (MS) patients experience symptoms of fatigue. MS-related fatigue is only partly explained by structural (lesions and atrophy) and functional (brain activation and conventional static functional connectivity) brain properties. OBJECTIVES: To investigate the relationship of dynamic functional connectivity (dFC) with fatigue in MS patients and to compare dFC with commonly used clinical and MRI parameters. METHODS: In 35 relapsing-remitting MS patients (age: 42.83 years, female/male: 20/15, disease duration: 11 years) and 19 healthy controls (HCs) (age: 41.38 years, female/male: 11/8), fatigue was measured using the CIS-20r questionnaire at baseline and at 6-month follow-up. All subjects underwent structural and resting-state functional MRI at baseline. Global static functional connectivity (sFC) and dynamic functional connectivity (dFC) were calculated. dFC was assessed using a sliding-window approach by calculating the summed difference (diff) and coefficient of variation (cv) across windows. Moreover, regional connectivity between regions previously associated with fatigue in MS was estimated (i.e. basal ganglia and regions of the Default Mode Network (DMN): medial prefrontal, posterior cingulate and precuneal cortices). Hierarchical regression analyses were performed with forward selection to identify the most important correlates of fatigue at baseline. Results were not corrected for multiple testing due to the exploratory nature of the study. RESULTS: Patients were more fatigued than HCs at baseline (p = 0.001) and follow-up (p = 0.002) and fatigue in patients was stable over time (p = 0.213). Patients had significantly higher baseline global dFC than HCs, but no difference in basal ganglia-DMN dFC. In the regression model for baseline fatigue in patients, basal ganglia-DMN dFC-cv (standardized ß = -0.353) explained 12.5% additional variance on top of EDSS (p = 0.032). Post-hoc analysis revealed higher basal ganglia-DMN dFC-cv in non-fatigued patients compared to healthy controls (p = 0.013), whereas fatigued patients and healthy controls showed similar basal ganglia-DMN dFC. CONCLUSIONS: Less dynamic connectivity between the basal ganglia and the cortex is associated with greater fatigue in MS patients, independent of disability status. Within patients, lower dynamics of these connections could relate to lower efficiency and increased fatigue. Increased dynamics in non-fatigued patients compared to healthy controls might represent a network organization that protects against fatigue or signal early network dysfunction.

Functional connectivity between resting-state networks reflects decline in executive function in Parkinson's disease: A longitudinal fMRI study.

Deficits in cognitive functioning are a common yet poorly understood symptom in Parkinson's disease (PD). Recent studies have highlighted the importance of (dynamic) interactions between resting-state networks for cognition, which remains understudied in PD. We investigated how altered (dynamic) functional interactions between brain networks relate to cognitive dysfunction in PD patients. In this fMRI study, 50 PD patients (mean age 65.5 years ± 6.27) on dopaminergic medication were studied cross-sectionally, and of this cohort 31 PD patients were studied longitudinally. MRI imaging and neuropsychological testing was performed at two time points, with a follow-up duration of approximately three years. Functional connectivity within and between seven resting-state networks was calculated (both statically and dynamically) and correlated with four neuropsychological test scores; a combined score of (four) executive tasks, a motor perseveration, memory, and category fluency task. Cognitive dysfunction was determined based on a longitudinal sample of age-matched healthy controls (n = 13). PD patients showed dysfunction on six out of seven cognitive tasks when compared to healthy controls. Severity of executive dysfunction was correlated with higher static and lower dynamic functional connectivity between deep gray matter regions and the frontoparietal network (DGM-FPN). Over time, declining executive function was related to increasing static DGM-FPN connectivity, together with changes of connectivity involving the dorsal attention network (amongst others with the ventral attention network). Static functional connectivity between the ventral and dorsal attention network correlated with motor perseveration. Our findings demonstrate that in PD patients, dysfunctional communication between (i) subcortical, fronto-parietal and attention networks mostly underlies worsening of executive functioning, (ii) attention networks are involved in motor perseveration.

Lipid-suppressed and tissue-fraction corrected metabolic distributions in human central brain structures using 2D 1 H magnetic resonance spectroscopic imaging at 7 T.

INTRODUCTION: Magnetic resonance spectroscopic imaging (MRSI) has the potential to add a layer of understanding of the neurobiological mechanisms underlying brain diseases, disease progression, and treatment efficacy. Limitations related to metabolite fitting of low signal-to-noise ratios data, signal variations due to partial-volume effects, acquisition and extracranial lipid artifacts, along with clinically relevant aspects such as scan time constraints, are among the challenges associated with in vivo MRSI. METHODS: The aim of this work was to address some of these factors and to develop an acquisition, reconstruction, and postprocessing pipeline to derive lipid-suppressed metabolite values of central brain structures based on free-induction decay measurements made using a 7 T MR scanner. Anatomical images were used to perform high-resolution (1 mm3 ) partial-volume correction to account for gray matter, white matter (WM), and cerebral-spinal fluid signal contributions. Implementation of automatic quality control thresholds and normalization of metabolic maps from 23 subjects to the Montreal Neurological Institute (MNI) standard atlas facilitated the creation of high-resolution average metabolite maps of several clinically relevant metabolites in central brain regions, while accounting for macromolecular distributions. Partial-volume correction improved the delineation of deep brain nuclei. We report average metabolite values including glutamate + glutamine (Glx), glycerophosphocholine, choline and phosphocholine (tCho), (phospo)creatine, myo-inositol and glycine (mI-Gly), glutathione, N-acetyl-aspartyl glutamate(and glutamine), and N-acetyl-aspartate in the basal ganglia, central WM (thalamic radiation, corpus callosum) as well as insular cortex and intracalcarine sulcus. CONCLUSION: MNI-registered average metabolite maps facilitate group-based analysis, thus offering the possibility to mitigate uncertainty in variable MRSI data.