Session 10 - Structural connectivity and the connectome Flashcards
(64 cards)
1
Q
Grey matter
A
- contains neuron cell bodies and extensive dendritic branching
- grey because of the clustering of cell bodies and dendrites
2
Q
White matter
A
- contains mainly axons (output fibers of neurons) –> myelinated
- appear white due to myelin (fatty insulation) produced by glial cells (esp. oligodendrocytes)
3
Q
Myelin
A
- insulates axons and enhaces signal transmission, making the tissue appear white in brain preparations and imaging
4
Q
Cortex and beneath
A
- outer layer of the brain
- composed of grey matter
Beneath cortex:
- deep regions filled with white matter –> axons extend across different parts of the brain to facilitate communication
- axons often do not simply connect to nearby neurons but instead travel long distances across the brain to form extensive communication pathways
–> difference between short local connections and long-range projections captured the essence of grey and white matter distinctions
5
Q
Brain as a structural network
A
white matter = wiring system of the brain
- connects different regions into a complex and functional network
–> may appear disordered in some imaging techniques BUT are highly organised into fiber tracts or fascicles
- beneath cortex: dense array of white fiber tracts that form the physical substrate of the brain’s communication system
6
Q
Major Fiber Tract categories
A
- three primary categories: projection fibers, association fibers, commissural fibers
7
Q
Projection fibers
A
- connect the cortex with subcortical regions or the body
Examples:
- optic radiation which originates in the lateral geniculate nucleus (LGN) of the thalamus and projects to V1
–> relays visual input from the eyes to the cortex
- Corticospinal tract which originates in the motor and somatosensory cortex and travels downward through the brain into the spinal cord
–> carries motor commands from the brain to the body and brings sensory feedback back to the brain
8
Q
Association fibers
A
- connect regions within the same hemisphere
Example:
- arcuate fasciculus which links Broca’s area, involved in speech production with Wernicke’s area, involved in language comprehension
–> fMRI studies show both areas activating strongly during auditory tasks, confirming the functional importance of their connection
9
Q
Commissural fibers
A
- connect corresponding regions of the left and right hemispheres
–> forceps major links the occipital lobes through the posterior corpus callous, while the forceps minor connects the frontal lobes through the anterior corpus callosum
10
Q
Standardised tract identification
A
- modern neuroimaging leverages automated tools like Extract to identify and reconstrut 42 standard white matter tracts
- reconstructions are based on origin and target regions of fiber pathways
- are determined through a fully automated, reproducible pipeline –> ensure consitency
- tracts included represent the most functionally significant and consistently observable pathways
Some tracts are excluded:
- very short U-fibers
- fine grained thalamic radiations
- several shorter and less consistently tracked tracts
–> due to small size or high variability across individuals
11
Q
Diffusion-weighted imaging
A
= a form of MRI that measure how water moves through the brain
- technique uses strong magnetic fields and radiofrequency pulses
- radio waves are sent into brain –> signal gets reflected (similar to echo) –> picked up by receiving coils
- magnetic gradients help encode spatial information, allowing the resulting singal to be mapped to specific brain regions
- presence or absence of directional/aligned fibers influence water diffusion
- in white matter: typically along a single dominant direction = anisotropic diffusion
- in ventricles: free diffusion of water in all directions = isotropic diffusion
- DWI makes diffusion patterns visible voxel-by-voxel
- even though DWI may initially appear noisy and indistinct, appropriate modelling allows detailed reconstruction of the underlying fiber architecture
11
Q
Measuring structural connectivity
A
- is inferred by analysing how water diffuses through tissue
- brain is composed largely of water and fat
- myelin: 75% fat
- water and fat repel each other so warer molecules do not diffuse through the myelin but rather along the direction of the axon fibers –> directional and anisotropic dissusion is basis for imaging white matter
12
Q
DWI: Example - Corpus Callosum
A
- one of the brain’s most heavily myelinated strcutures, region where white matter is especially prominent
- high resolution microscopic images (as small as 50 micrometers)
- computer reconstructions show that axons in the CC are densely packes, myelinated and aligned in parallel, running in consistent directions across very small spaces
- at the scale of typical MRI voxels: most axons and their myelin sheaths are uniformly oriented
- this form of organisation is not present throughout the brain –> areas like grey matter or cerebrospinal fluid-filled ventricles (less mylein, fewer aligned fibers) do not show/only minimal directional alignment
–> influences water diffusion
13
Q
Model water diffusion: DTI
A
- mathematical techniques applied to estiamte diffusion in each voxel
- most widely used approaches is Diffusion Tensor Imaging (DTI)
- water diffusion is represented by a tensor (= mathematical object (3x3 matrix) that describes how diffusion occurs in three dimensions)
- tensor takes form of ellipsoid (= stretched/flattened surface) –> illustrates how water molecules move in various directions within a voxel
- shape and orientation of ellipsoid reveal dominant directionality of diffusion
- three values = estimated from DW MRI data
- can be fit to each voxel based on these observations
Two essential pieces of information:
1. quatifies how strongly diffusion is oriented (anisotropy)
2. eigenvectors (= descriptions of direction and magnitude): identify the dominant direction of diffusion within a voxel
–> direction generally correpsonds to the orientation of axon bundles in white matter –> results in the tensor revealing both whether diffusion is directional and the likely orientation of the underlyinf fiber pathways
- this (tensor) information forms the basis for advanced tractography and structural connectivity mapping
14
Q
DTI: fractional anisotropy (FA)
A
- ranges from 0 (= completely isotropic diffusion, equal in all directions) to 1 (= highly directional, anisotropic)
- high FA values typically reflect healthy, well-meylinated white matter
- can be visualised in colour FA maps –> primary diffusion direction for each voxel and reveal anatomical structure based on spatial FA patterns
15
Q
DTI: limitations
A
- nerve fibers don’t travel in neatly seprarated bundles
–> weave, overlap and cross - DTI only assumes one dominant fiber direction per voxel
- in complex regions like the centrum semiovale, major tract intersect –> the superior longitudinal fasciculus (SLF) and the corticospinal tract (CST)
–> when multiple fiber populations aer present within a voxel, model may fail to capture their true configuration and incorrectly report a low FA value (even though it is highly directional)
16
Q
Reconstructing Fiber tracts (tractography)
A
Streamline tractography = used to create virtual model of fiber tracts
- builds virtual paths by moving from voxel to voxel, following the estimated direction of diffusion through the brain
- process stops when certain criteria are met
–> drop in FA or sharp change in curvature
- involves interpolation across neigbouring voxels –> for each point along a streamline, algorithm considers not only the main diffusion direction in the voxel but also the directions in adjacent voxels
–> helps smooth out the resulting streamlines, producing more gradual and anatomically plausible trajectories
Continuous assignmet = algorithms typically extend streamlines unless the next step would create a biologically implausible turn
- axons do not make abrupt turns –> curvature thresholds are applied to prevent unrealisitic pathway reconstructions
- ensures that the modelled streamlines reflect the smooth, continuous paths of actual white matter fibers
BUT
- indirect method
- streamlines are algorithmic constructs based on models, data quality, parameter settings –> can miss actual anatomical connections, generate false ones, perform poorly in regions with compley geometry (fiber crossings, branchings, loops)
- streamlines are model, not literal wires in the brain –> brain could prodcue different connectomes if analysed differently
17
Q
Extracting and quatifying fiber tracts
A
- identify specific anatomical tracts within this complex web of conections
- step-by-step process to reconstruct meaningful fiber tracts (DWI)
- Whole-brain tractogram generation:
- first, exclude voxel with low directional diffusion (low FA) and prevent streamlines form making sharp and biologically implausible turns
- place mutiple seed points in every voxel (16+) and allow streamlines to propagate through plausible regions and discard those that curve too sharply or cross low-FA areas –> result is a whole-brain tractogram, a dense set of streamlines representing possible white matter pathways - Isolating specifc fiber tracts:
- to extract a specific tract (like forceps major)
- apply anatomical constraints
–> seed masks specify where streamlines must begin or pass through (eg visual areas V1-V4 for forceps major)
–> inclusion masks specify other regions a streamline must pass through to be kept
–> exclusion masks remove any streamline that enters regions inconsistent with the known anatomy of the tract
- for the forceps major, streamlines must begin in the occipital lobe of one hemisphere, pass through the corpus callosum and end in the occipital lobe of the opposite hemisphere
- Why this matters: measure key properties
- one key measure is fractional anisotropy (FA), which reflects how stronglym water diffusion is directionally organised
- by averaging FA values across all voxels within a tract, we get a single number –> integrity score for that tract
- everyone has same basic tracts (like forceps major) but integrity can vary
- higher average FA suggests stronger, more coherent connections, which may relate to brain function or health
18
Q
Advanced models and tractography strategies
A
- to overcome limitations of DTI –> more advances models like Constrained Spherical Deconvolution (CSD) are used
- can detect multiple fiber directions in a single voxel by estimating a fiber orientation distribution function (fODF) -> looks like a spiky ball showing probable fiber orientations
- CSD works by defining a response function from a voxel known to contain a single fiber population (usually in the corpus callosum)
- measured signal is assumed to be a mixture of this response function in different directions
- deconvolution gives the model estimates of directional composition of voxel
- CSD requires richer data
–> includes measuring diffusion in many directions (typically 60-128) using higher b-values (stronger diffusion weighting) and conducting longer scan sessions with high-performance MRI scanners (advanced scanner hardware is needed, 3T MRI with stronger gradients)
19
Q
local and global tractography
A
Local tractography
- builds connections step-by-step, moving from voxel to voxel using nearby data
- fast and common, but small errors can build up –> especially tricky in areas like noisy or low FA areas
Global tractography
- big picture approach –> looking at whole brain at once to find most likely overall set of connections
- better handeling of noise and complexity and more reliability in hard-to-track areas
-BUT: slower pace, use of more computing power, not widery available in standard tools
- this distinction helps explain why tractography results can differ –> not only between software but also depending on tracking method
20
Q
Challenges/limitations
A
- crossing fibers are better detected now but remain difficult to fully solve
- long-range tracts can pick up errors as they stretch across the brain
- deterministic tracking (following one path at a time) may look precise but can be overly confident
- probabilistic tracking (testing many possible paths) add nuance:
–> pro: captures uncertainty, reveals weaker or less direct connections
–> con: more complex, harder to interpret, computationally demanding - DWI does not show direction –> we can see connection but not direction of signal transmission (listener vs talker)
21
Q
Ground truth and method validation
A
- tractography is indirect method –> researchers validate models using animal studies
- anterograde tracers like PHA-L show where neurons send their outputs
- retrograde tracers like CTb reveal where inputs come from
- viral tracers (including rabies, herpes viruses) can even cross synapses
–> methods provide highly accurate connectivity data but are invasive and not usable in humans - CoCoMad database (non human primates) confirms that many brain connections are bidirectional, although one direction often dominates
22
Q
Which tracts are most reliable?
A
- some tracts are more consistently reconstrcuted across studies
- association fibers: superior longitudinal fasciculus (SLF), inferior fronto-occipital fasciculus (IFOF), uncinate fasciculus
- projection fibers: corticospinal tract, thalamocortical pathways
- to measure strength or integrity: several metrics
–> FA reflects directional diffusion
–> number of streamlines (nOS) counts how many virtual oaths connect two regions
–> streamline volume density adjusts for difference in region size (makes comparisons more accurate) - each metric highlights different aspects of connectivity and is influenced by analysis choice
23
Q
From tracts to connectivity
A
- instead of isolation –> often how defined brain regions are interconnected
- one approach: divide the brain into parcels such as the 360-region map used in the Human Connectome Project
–> allows for detailed analysis of which regions are connected and how strongly
–> connectivity matrix is created where rows and columns represent individual brain regions
–> each cell indicates whether a connection exists and how strong it is (based on metrics like number of streamlines or mean FA) - matrices often reveal that there are more connections within hemispheres (association fibers) than between hemispheres (commissural fibers) –> by analysing such matrices, researchers can identify modules and patterns tha reflect the brain’ structural organisation
24
The brain as an incredibly efficient network
- aim: examine the organisational principles that govern all the connections in the brain
- despite constraints like spatial geometry, wiring cost, energy consumption--> architecture allows for highly effective communication across billions of neurons
- how? Connectome
25
The connectome: definition and concept
= totality and organisational structure of all connections in the brain/entire nervous system
- encompasses the entire set of long-range fiber pathways that connect different brain regions of the cortex
- fibers are from organised bundles that enable communication between brain areas, are embedded within the white matter (beneath cortical sheet)
- allows for the study of connectomes at micro, meso and macro scales
- connections forming the connectome are white matter fiber tracts and these tracts are myelinated --> serves accelerated signal conduction, insulation improves speed and efficiency of communication between distant brain regions
26
Connectome: history
- neologism
- introduced independently in 2005 by two researchers: Patrick Hagman and Olaf Sporns --> used term to describe the map of neural connections, including the underlying principles that govern their organisation
27
Network components of the connectome
- connectome represents the brain's network map and is composed of two essential elements: nodes and edges
nodes
- or network nodes
- individual elements that are connected within the network
- usually defined as brain regions (in brain networks)
- also possible to define nodes at a more fine-grained level --> individual voxels can serve as nodes (even though voxels are technically defined an not biologically meaningful per se)
edges
- or network edges/connections
- are links between nodes
- are based on white matter connections (in structural connectome) = fiber tracts
- connectome can also be defined on functional coupling
--> edges represent transient, functional relatioships between brain areas (typically derived from fMRI data)
--> functional connectome
- despite flexibility in node and edge definitions, standard approach:
--> nodes are brain regions, based on anatomical or literature-based parcellations
--> edges are defined as structural fiber connections
- allows for consistency in research and is the model most commonly used in studies
28
Levels of Connectome Analysis
1. Mirco-level connectome
2. Meso-level connectome
3. Macro-level connectome
29
Micro-level connectome
- at microscale: nodes could be defined as individual neurons (synaptic contacts)
- edges would then represent synaptic interaction in a detailed circuit diagram
- due to current technical and data limitations --> not yet possible in humans or even in most mammals
- only nervous system that has been fully reconstructed at this level is that of the small roundworm --> contains over 200 neurons (prove of concept)
30
Meso-level connectome
- analysis shifts to slightly larger structures --> like cortical columns in cerebral cortex
- columns encompass many neurons and act as semi-autonomous funcitonal units with distinct inputs and outputs
- are more tractable than individual neurons BUT is methodologically and financially challenging
31
Macro-level connectome
- especially in cognitive neuroscience
- nodes are defined as larger anatomical units such as brain regions
--> voxels (technically defined without inherent biological meaning)
--> named brain areas, such as Brodmann areas or updated corical parcellations
- larger and less precise
- regions are often functionally coherent and can be treated as distinct porcessing units in network analyses
- brain is modelled as a network: nodes represent brain areas and edges represent connection (structural fiber tracts)
32
Structural connectivity reconstruction using diffusion imaging
- connections between brain regions/voxels can be represented non-invasively through advanced neuroimaging techniques--> most widely used approach = diffusion-weighted imaging (DWI)
Reconstruction
- network is assumed to exist and is being modelled for analytical purposes (reling on imaging data)
Workflow:
1. collecting MRI data
- two distinct types ar required: T1 and diffusion-weighted scans
- T1: provide high-resolution anantomical detail, allowing for the delineation of individual brain structures and regions
- diffusion weighted scans: many are needed for higher reliablity of estimates, often require quantitative mesures of the brain's white matter fiber pathways
2. Further analyses:
- tractogram based on dw-scans --> applying streamline reconstruction algorithms that estimate the likely paths of white matter fiber tracts throughout the brain (diffusion patterns of water molecules)
- T1-scan data is used to define and segment various brain regions
--> overlaying anatomical parcellations and tractogram allows to assess whether fiber streamlines pass between any given pair of brain areas --> structurally connected
- number of streamlines linking them can serve as an indicator of connection strength
Example:
- superior frontal gyri: heavily interconnected via corpus callosum
- orbitofrontal cortex and posterior parietal region are only connected by a single streamline --> too sparse to support a meaningful structural connection
3. Connectivity matrix
- pairwise evaluation is repeated for the whole brain
- can be a simple 0-1-matrix (binary)
33
Insights from connectivity matrix
- structural organisation of the brain
- group connectome = aggregates data from several hundred individuals, producing a best average or consensus model of brain connectivity
- organised as a region-by-region matrix --> each entry represents whether two regions are structurally connected
- first half of rows/columns represent left half, second represents right half
- allows certain large-scale patterns to become visually apperent
--> upper-left and lower-left quadrants (intra-hemispheric connections, association connections): contain a high density of connections
--> upper-right and lower-right quadrants (inter-hemispheric connections, commissural connections) show relatively sparse connectivity
- indicates that intra-hemispheric connections (association connections) are more numerous and stronger than inter-hemispheric (commissural connections)
- empty main diagonal: does not mean that there are no connections --> simply reflects that DWI cannot capture them
- all direction are reciprocal and bidirectional--> limitation of method
34
Insight from connectivity matrix: part2
- ordering of brain regions in matrix follows a hierarchical gradient from unimodal to multimodal or transmodal areas
--> unimodal = are primarily involved in a single sensory modality, such as purely visual, auditory, somatosensory or motor areas
--> multimodal or transmodal = integrate input from multiple sensory modalities and are associated with higher-order cognitive functions
- connectivity is stronger among unimodal regions compared to association ares --> suggests that regions involved in basic sensory or motor processing are more densely interconnected than those involved in higher-level integrative processing
homotopic connectivity = nature of inter-hemispheric/commissural connections --> often link homotopic areas (=functionally and anatomically corresponding regions in the opposite hemispheres) --> V1 in one hemisphere is typically connected to V1 in opposite hemisphere
35
Connectivity matrix insights: summary
- there are more connection within each hemisphere than between them
- commissural connections often link homotopic regions
- unimodal areas exhibit stronger internal connectivity compared to multimodal or association areas
--> highlight the rich structual architecture of the brain and demonstrate how much can be inferred form a well-constructed connectome
36
Structual organisation of the brain network
- three foundational observations emerge from analysis
1. density of the network can be assessed
- calculating the number of actual connections in the network relative to the number of theoretically possible connections
--> corresponds to the ratio of realised connections to potential connections
- brain network is sparsely connected --> arises from the fact that only a small proportion of all possible connections are actually present
- commonly observed density value is approx. 8% --> only 8% of all conceivable connections within the brain are realised
2. significant inter-indidivual differences
- the absence of a connection in one individual strongly predicts the absence in another individual
- BUT presence of a connection does not guarantee presence in another individual
--> challenges the intuitive assumption that a consistent set of brain connections should exist across all humans due to underlying genetic and biological programming
- organisational principle = architectural logic underpinning the brain network
--> exhibits a remarkable degree of consistency across individuals
--> prominent aspect of shared organisation is the uneven distribution of connections accross different brain regions
- exceptionally high connectivity in some regions --> hyperconnectivity measured by nodal degree
nodal degree = the total number of connections it maintains with other regions
- can be calculated by binarising the connectivity matric and summing values in each row/column
37
Biological constraints and the effiency of brain network architecture
1. Biological constraints related to physical space
- skull offers limited room for tissue and connective wiring
- solved by gyrification and sulcification
--> highly folded, allowing a larger cortical surface area to fit within the confined space of the skull --> maximisation of use of availabe space
- limitation of space might be an explanation for spare neural connectivity --> no space for fully connected system
2. Energetic constraints
- 2% of body mass --> consumes 20% of the body's total energy (glucose and oxygen)
- majority of energy used to maintain conditions necessary for neuronal signaling
--> restoring membrane potentials and maintaing synaptic ion gradients
--> even if additional space was available, the metabolic costs would be prohibitive
--> evolutionary necessity of efficient organisation
(anticipated principle of efficient network organisation by Santiago Ramón y Cajal)
38
Organisational principles of the brain network
Three core principles:
1. the brain is structured as an efficient small-world network
2. The brain network is organised modularly
3. the brain network contains central hub regions, which serve as the backbone for information processing
Wiring rules:
1. Connect to your neighbours
- most connections are short range
- probability of connection decreases with spatial distance between regions (geometric constraint)
- a few long range connections are permitted, giving rise to the classic small world structure
2. Connect to popular nodes
- neurons or brain areas are more likely to connecto to already well-connected regions
- leads to emergence of hubs --> rich club structure
- network simulations using these rules can reproduce much of the real-world topology --> from there, Hebbian learning to fine-tune the system
38
Small-world properties of brain networks
- concept originated from network science and refers to a type of network that is extremely efficient in its structure and function
- two key parameters:
1. Clustering coefficient
2. Path length
39
Small-world properties of brain networks: Path length
- refers to how many steps on average are required to travel from one node to another
- in brain networks, there are significantly fewer actual connections than theoretically possible --> direct links between all brain regions are rare
--> communication with brain region that is not directly connected must be relayed through intermediary regions
- the fewer intermediary regions are passed the shorter the path and the more efficient the network
40
Small-world properties of brain networks: Clustering coefficient
- measure the degree of local interconnectivity within a network
--> among three regions two are connected, how likely is it that the this regions is also connected to the others? the more 'closed triangles' present, the higher the clustering coefficient
- a high clustering coefficient indicates strong local connecitivity
41
Small world: Idealised network structures
Ideal types of networks:
1. Lattice network:
- each region is connected to its direct neighbours
- ensures perfect local clustering but results in long path lengths because information must traverse many intermediate nodes
2. Random network:
- distributes connections randomly among nodes
- typically results in short path lengths --> statistically, fewer steps are needed between nodes
- BUT local clustering is sacrificed, lack of intermediate connections
3. Combining structures: the small-world network
- combination of lattice-like regularity and random shortuts
- retains high local clustering while incorporating strategically placed long-range connections to reduce path lengths
- named like this because any node in the network can be reaches from any other through only a few intermediate steps --> most efficients network configuration
42
Empirical evidence for small-world networks in the brain
- empirical brain networks can be reconstructed from imaging data
- two jey metrics: average path length and clustering coefficient
- to facilitate comparison, synthetic (null) models are created --> either idea lattice or randomly connected networks
--> deviation quantified by Small World Propensity (SWP)
SWP
= metric calculated by squaring the deviation of the empirical clustering coefficient from the perfect lattice/random network, substracted from 1
--> the closer to 1 the more the empirical networks resembles a perfect small-world network (>0.6)
- applied in multiple studies and species --> both micro- and macro-level analyses show consistently small-world properties
- brain network is highly efficient both at the local level (high clustering) and the global level (short paths lengths)
43
Modularity in brain networks
- brain networks exhibit modularity = structure in which subsets of regions (modules) are more strongly interconnected internally than with other modules
44
Functional connectivity modules
- modularity is evident in functional intrinsic connectivity networks where certain brain areas show synchronsied activity
- functionally connected modules are considered foundational systems of cognition
--> strong internal activity (underpinned by underlying structural fiber paths)
45
Examples of brain modules
- top-down view of the brain can show modular organisation more intuitively
Default mode network
- oven active during rest and internal thought
Action mode network
- associated with goal-directed activity
46
Central hub regions in brain networks
- hub regions are highly connected and critical for information integration across network
- have significantly more connections than average
- bridge different network modules
- facilitate global integration of infomration while maintaining modular segregation and enable efficient communication across the entire network
- maintaining these is metabolically expensive --> require structural resources (axons, myelin) and more energy (increased glucose and oxygen)
--> BUT essential to brain function
47
Identifying hubs: key metrics
various metrics used to identify hub regions
1. nodal degree = total number of connections a node has, regardless of where they go
2. nodal path length = average number of steps required to reach all other nodes from a given node --> nodes with lower values are better integrated
3. participation coefficient = indicates how connections of a region are distributed across different modules --> high participation coefficient suggests a connector hub that links multple modules
4. Within-modules z-score = measures how many connections a region has within its own module --> high score suggests a provincial hub that is central within its own modules but less connected to others
5. betweenness centrality = assesses how often a nodes lies on the shortest paths between other nodes
- regions with high betweenness centrality are crucial transit points for information flow and are thus strong hub candidates
- are often combined
48
Prominent brain hubs
Consistently identified hub regions
- Precuneus = major hub within the DMN
- posterior cingulate cortex = closely associated with the precuneus in the DMN
- superior frontal gyrus
- lateral occipital cortex = despite role in visual processing, functions as hub
--> play central roles in coordinating information processing across the brain
49
Concept of rich club
- hub regions form a collective (= interconnected group that can be considered a distinct module) --> forms a higher-level organisational unit known as the rich club
Rich club
- derived from two characteristics:
1. Richness in connectivity: densly connected and high number of connections
2. Club-like structure: rich club structures are even more closely connected with each other than one would statistcially expect --> tight knit network/clique of highly connected brain areas
--> central importance stemps from their ability to support a significant portion of the brain's total information traffic
- specific regions that comprise the rich club may vary depending on how the brain network is defined
- all rich club regions are hub regions but not all hub regions are rich club regions
--> rich club as subset of hub regions with particularly strong mutual connectivity
50
Cross-species presence of the rich club
- appears to be universal organisational principle of neural networks and isn't unique to humans
- found across different species, different levels of biological complexity
51
developmental insights
- studies using fetal and pre-term MRI reveal the basic layout of the brain's network is in place as early as gestational week 20 --> halfway trhough pregnancy
- connectome is already modular --> exhibits a small-world organisation and includes a rich club of highly connected hub regions
52
maturation and reorganisation
- development does not stop at birth
- brain networks continue to evolve throughout infancy, childhood, adolescence and well into adulthood
- two major trends
1. synaptic pruning = reduces local connections, leading to lower clustering and greater efficiency
2. long-range connections strengthen = reinforcing the hub structure, rich get richer
--> changes may enhance efficiency
--> BUT may come at cost of flexibility --> more stable, role-specific architecture
53
Aging: reverse development?
- prominent hypothesis suggests that aging is like reverse development--> what matures last, declines first
--> rich club structure becomes even more enhanced with age
- overall connectivity weakens
- implications for cognitive aging, reserve and resilience are not fully understood
54
Funcional significance and biological costs
- rich club possesses an extremely high capacity for information processing --> BUT biological cost
- significant amount of biological material is required for the development and maintenance of the rich club
- demands a disproportionately large share of the brain's energy resources
55
classification of neural connections
- based on rich club: three types of connections can be distinguished
1. Rich club connections = connect two regions within the rich club
2. feeder connections = connect peripheral regions to rich club regions (or vice versa)
3. local connections = connect peripheral regions with each other, without involving the rich club
56
length and resource use of rich club connections
correlation betwen length and rich club membership:
- long connections (greater than 90mm) are more liekly to be rich club connections
- short connections (less than 30mm) are typically local and connect peripheral regions
- despite fewer in number --> rich club connections require the most biological material and have the greatest length --> consume disproportionate share
57
communication paths through the brain
- rich club plays a cruical role in information transmission--> especially for long-distance communication
1. short paths: often use local connections and do not involve the rich club
2. medium-length paths: typically involve feeder conections that link peripheral areas to the rich club and back
3. long paths: traverse through the rich club via rich club connections
58
The connectome and cognition
- brain networks are not just anatomical --> deeply funcitonal
- major focus on modularity and hub regions (particularly connector hubs)
Cognitive architecture
- cognition itself is modular --> a single task involves several subprocesses
1. selective attention to relevant stimuli
2. perception in a specific sensory modality
3. memory maintenance
4. executive control to follow task rules
- functions may involve distinct brain modules
- higher cognitive tasks require integration (--> connector hub time)
- the more complex the task, the more connector hub activity increases
- confirms the autonomy of network modules and the central role of connector hubs in integration
Individual differences in cognition
- connector hub are not just important for performance, they also predict individual cognitive ability
- people with more diversely connected hubs tend to have more modular and efficient networks
- hubs actively shape the connectivity of their neighbours, supporting both local specialisation and global integration
59
The brain network - structure and significance
- human brain is highly complex network
- consists of ~86 billion neurons which connect with one another via their axons to form fiber bundles
- bundles interlink all brain regions, resulting in an integrated overall network referred to as the connectome --> organised efficiently
--> structure is modular: composed of distinct sub-networks or modules that specialise in particular functions
--> at the same time remains centrally connected: hubs in the so called rich club (highly interconnected that facilitate efficient communication across the entire brain)
- brain network exhibits small-world architecture
--> combines high local clustering of connections with relatively short path lengths between any two nodes, allowing for both specialised processing and global integration of information
- may also offer insights into origins and mechanisms of certain mental disorders
60
The connectome and brain disorders
- new insights into psychiatric and neurological disorders by comparing clinical groups to neurotypical controls
Small-World disruptions
- ADHD shows increased clustering --> suggests a developmental lag: the nework resembles an earlier stage of brain maturation
- Schizophrenia is characterised by more random connections and shorter communication paths, possibly leading to excessive crosstalk between modules --> explanation for symptoms like disorganised thinking or delusions
Rich-Club vulnerability
- resilience to random damage: simulations studies show that the brain is robust against random failures unless it hits a rich-club node --> unlikely to target them
- targeted disruptions: disorders might resemble targeted attacks
--> in schizophrenia, rich-club connectivity is impaired and similar patterns are seen in unaffected first-degree relatives (genetic component)
Transdiagnostic impairments
- cross-disorder analysis showed that the same connections are often affected across many disorders
- PTSD, OCD, Bipolar disorder, depression, autism spectrum disorder, mild cognitive impairment, Alzheimer's, ALS, ADHD
- commonly impaired connections are most likely rich-club connections or feeder connections (connect hubs to periphery)
61
Future directions
- longitudinal studies --> in Alzheimer's for example: grey matter atrophy and amyloid depositions are more pronounces in rich-club regions
--> time line is unclear: could help to understand spread of disease-related disconnection
62
Why the connectome matters
- brain is a complex network: efficiently wired, dynamically organised, functionally modular
- modern neuroimaging techniques allow us to reconstruct whole brain connectomes
- across development, cognition, disease: brain's network architecture proves critical (small-world structure, modularity, rich-club hubs)
- systems-level view opens up powerful new avenues in neuroscience: for understanding how we grow, how we think and what goes wrong in the brain
