1. Network hubs in the human brain
정미르

3. Introduction
Brain: anatomically differentiated, functionally specialized (early times) ⇒ Specialization × fully account most functions
Integrative process, dynamic interaction underpin cognitive processes
What’s the neural substrate of integration?
1. Neural communication: synchronization, information flow
2. Specific brain regions & anatomical connections: complex & diverse response, ↑level hierarchy, confluence zones

4. Introduction Approach: ‘complex network’
Neural system: graphs, networks
Network map, connectome: structural basis for dynamic interactions
1. unravel its architecture
2. explain how structural network topology shape & modulate function
Connectome: diverse; modularity (specialization)+communication (integration)
Network hub ensure integration: high degree connectivity, central placement

5. 1. Methodological Aspects: Detection and Classification of Hubs in Brain Networks

6. Brain Network as Graphs
Nodes- neuronal elements
Edges- their interconnections
Pair-wise couplings: connection matrix
Arrangement: topology
Graph theory offer data-driven measures to characterize network topology
Identifies network elements having strong influence on the global network function

7. Hub Detection
Network hub: nodes that make strong contributions to global network function
Detected by…
1. Node degree (degree centrality): # of edges maintained by each node
2. Eigenvector centrality (pagerank centrality): favor nodes that connect to other highly central nodes
3. Closeness centrality: average distance btw a given node and the rest of the network
4. Betweenness centrality: # of short communication paths a node (or edge) participates in
5. Vulnerability: node deletion impact on global network communication
6. Dynamic importance: synchronization; graph metrics before & after deletion
Rankings of nodes: highly correlated ⇒ aggregate rankings
4&5: identify central edges

8. Hub Detection
Network community(=module): nodes more densely linked among each other than with nodes in other communities
Participation coefficient: differentiates provincial hubs (single module) from connector hubs (multiple modules)
Peripheral nodes: low-degree, connects within own module, ↓participation coefficient
Hub nodes highly interconnected than by chance?
Rich club: collectives of high-degree nodes and their interconnecting edges ⇒ facilitate mutual interactions
Structural core: recursive pruning of nodes of increasing degree ⇒ resilient nodes (densely interconnected)

9. Structural Network Functional Network
anatomical connectivity: Relatively stable (sec~min); plasticity (h~day)
Edges: physical links; infrastructure for neuronal signaling & communication
Derived from statistical descriptions of time series data
fMRI ⇒ linear (Pearson) cross-correlations
Time dependent, modulated by stimuli and task context, non- stationary fluctuations even at rest
Edges: × represent anatomical connections
Links many structurally unconnected node pairs, prone to transitivity ⇒ over-connection & high clustering
Reflection of signaling & communication unfolded in structural network
Define Hubs based on network’s community structure

10. 2. Empirical Results: Candidate Hubs in the Structural and Functional Connectome

11. Structural Hubs
Betweenness centrality: dorsal superior frontal(I), precuneus(II), occipital gyrus(III)
Group-averaged: Precuneus(a), anterior(b) & posterior cingulate(c), superior frontal(d), dorsolateral prefrontal(e), insular(f), occipital(g), temporual gyrus(h)
Other measures: structural vulnerability, node degree, multiple centrality metrics rankings (medial parietal, frontal & insular)
Their Integrative & diversive properties ←central embedding within connection topology
Functional properties are shaped by ‘connectional fingerprint’

12. Structural Hubs
Also maintain ↑# of anatomical connections among each other
Hub regions: more densely interconnected (wrt degree alone)
⇒ rich club (densely interconneted core)
1. Inter-hub communication robustness↑
2. Promote efficient communication & integration across the brain
Other aspects for classification of hubs as rich
Structural network hubs & connections…
1. ↑wiring volume, ↑white matter (efficiency↑, cost↑)
⇒ Direct communication path, transmission delay↓, robustness↑
2. metabolic active
3. complex cellular & microcircuit properties
⇒ Shaped by trade-offs btw wiring cost, spatial &metabolic constraint & efficiency
Location of hubs: consistent
High-degree regions: parietal, frontal, insular cortices
Network hubs: universal feature of connectome organization

13. Functional Hubs
Mearuse density (concentration) of local & global network functional connectivity
Functional interactions: precuneus, (posterior) cingulate gyrus, inferior parietal, ventromedial frontal cortex
⇒ significant overlap w/ default mode network
Recent approach: characterization of functional heterogeneity
1. Assess coactivation levels from various cognitive tasks
2. Participation of hubs in multiple functional domains
3. examination of functional paths layout within the network’s
functional connectivity pattern
:‘step-wise connectivity’ approach (unimodal→higher-order)
Multimodal & Functional hubs: superior parietal, superior frontal cortex,
cingulate gyrus, anterior insula (portion): part of resting-state network

14. Functional Hubs
Other approach: examine participation across multiple functional networks & level of overlap btw different functional domain
Primary regions (primary motor, visual & auditory): single or small # of functional networks
Putative hub regions (medial superior frontal, anterior cingulate, precuneus/posterior cingulate gyrus): multiple functional networks
Flexible network hubs: capacity to link & interact w/ diverse brain regions adaptively (+time & temporal variability)
☞frontoparietal, medial parietal & posterior cingulate
Detection based on graph analysis: pair-wise statistical relationships
(correlation coefficient btw recorded time series)
Identification depends on
1. Network edge estimation methodology
2. Graph metrics expressing ‘functional centrality’
Spatial overlap w/ default mode network: central role
⇔ due to local interactions within them &
functional degree biased by the network size

15. Individual Differences and Development of Hubs
connectivity profile & functional coupling level variation: intelligence, performance in different cognitive domains, interhemispheric integration, personality traits
Intelligence: medial parietal & prefrontal
Cognitive control & intellectual performance: frontal
Personality: medial parietal & cingulate
Functional connectivity topology (high density): heritable
Genetic influence
⇒structural integrity of long-range white matter tracts

16. Individual Differences and Development of Hubs
Structural hubs emerge early, but functionally immature, confined to primary visual & motor
Hub regions remain relatively stable; their interaction undergo developmental changes
1. ↑Functional connectivity among association areas
2. Spatially localized ⇒ globally distributed functional network through brain development
Sex-related difference in hormone (eg. ↑LH) level affect white matter brain connectivity
Genetic & environmental factors ⇒ individual variation on connectivity ⇒ cognition & behavior variation

17. Hubs in Brain Dysfunction
Abnormal connectivity & functioning: cognitive impairment ←computational network studies
Schizophrenia: frontal hub connectivity↓, disturbed rich club formation-‘long-standing dysconnectivity hypothesis’
Childhood-onset schizophrenia: disrupted modular architecture & disturbed connectivity in multimodal association cortex
Autism: altered intra- & intermodular connectivity of densely connected limbic, temporal & frontal regions
Alzheimer’s: medial parietal
Frontotemporal dementia (FTD): frontal

18. Hubs in Brain Dysfunction
Focal damage affect behavioral & cognitive functioning
Lesions at functional connector hubs: disruption of modular organization
Cognitive decline associated with white matter damage & ↓network integrity
Damage to long-distance connections ⇒ network function & cognitive disruption

19. Hubs in Brain Dysfunction
Reduced levels of consciousness related to disruption of hub connectivity
Metabolic activity↓ in parietal precuneus & posterior cingulate hub
Coma: random reorganization of functional hubs
Vegetative & minimally conscious: persisting functional connectivity level

20. 3. Conceptual Framework: Role of Hubs in Communication and Integration

21. Hubs and Network Communication
Neural hubs derive their influence from their strong participation in dynamic interactions due to neuronal signaling
Infer communication patterns focusing on layout of short paths & centrality of nodes (relative to these paths)
Structural & functional hubs play a central role in global brain communication
Hub regions may also render them potential neural ‘bottlenecks’ of information flow: set upper bounds for integration, chaining & serializing mental operations
Mental disorders ← structural & functional connectivity disturbances
Limitations⇒ × intrinsic to network models; reflect our ignorance & lack of data
1. capture only inter-regional projections,
× include local circuits (transformation & recording)
2. × fully predict dynamic patterns of communication (local firing rates of neurons,
Level of activity, external input, coherent phase relations, synaptic efficiency)
some nodes may preferentially engage in neural communication; others rarely or never
3. assume nodes connect along most efficient paths, are accessible, dominant criterion
Require detailed studies that track actual network paths of information flow

22. Hubs as Sources and Sinks
Some cortical hub regions maintain an unequal balance of incoming and outgoing projections
Receivers (sink): frontal & paracingulate cortex
Emitters (source): cingulate, entorhineal, insular cortex
※ Consistent w/ inferred directionality of functional interactions

23. Hub Connections
‘edge-centric’ perspective: focus influence of edge on network organization
Hub-hub & hub-(non-hub) edges: ↑proportion of long-distance connections
1. Short communication relays: transmission delay↓, interference & noise↓, faster synchronization
⇒ ‘Small world’; shortcuts × randomly placed; aggregate at hub nodes
2. efficient neural communication
3. robustness of inter-hub communication
Strong hub connections (in white matter): structural modules & functional resting-state networks ⇐ vulnerability

24. Hubs and Cross-modal Integration
Brain hubs & their connections: convergent structure for integration of information, forming putative substrate for functional ‘global workspace’
☞cognitive architecture in which segregated functional systems can share & integrate information ←neuronal interactions
Connective core hypothesis: interconnected hub regions topologically central offer an important substrate for cognitive integration: broadcasting, dynamic coupling, offering ‘arena for dynamic cooperation and competition’
Confluence (convergence) zones: hub regions across multiple functional domain overlapping area
↑involvement of hub nodes in intermodular connectivity

25. Hubs in Computational Models of Brain Dynamics
Enable ↑level functional diversity & synchronization
Central hub nodes engage on more variable or noisy dynamics ⇒ structural rewiring
Highly connected hub nodes & connections dominate system’s dynamical organization
: desynchronized ⇒ centrally synchronized dynamics
Scale-free architecture: ↑functional diversity
Hub nodes and edges lesions: most disruptive for network organization & functioning
Network hubs: loci of high variability & plasticity (+maintaining the cortical synchronization, modularity structure, functional dynamics)

26. Thank You for listening!
Any Questions?