Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron...
-
date post
19-Dec-2015 -
Category
Documents
-
view
219 -
download
0
Transcript of Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron...
![Page 1: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/1.jpg)
Small-World Brain Networks
Danielle Smith Bassett and Ed Bullmore
Written By:Gassan TabajahRon AdarGil Rapaport
![Page 2: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/2.jpg)
Motivation Many complex networks have a small-world
topology characterized by dense local clustering or cliquishness of connections between neighboring nodes yet a short path length between any (distant) pair of nodes due to the existence of relatively few long-range connections.
This is an attractive model for the organization of brain anatomical and functional networks because a small-world topology can support both segregated/specialized and distributed/integrated information processing.
Moreover, small-world networks are economical, tending to minimize wiring costs while supporting high dynamical complexity.
![Page 3: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/3.jpg)
Motivation The authors introduce some of the key
mathematical concepts in graph theory required for small-world analysis and review how these methods have been applied to quantification of cortical connectivity matrices derived from anatomical tract-tracing studies in the macaque monkey and the cat.
The evolution of small-world networks is discussed in terms of a selection pressure to deliver cost-effective information-processing systems.
![Page 4: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/4.jpg)
Motivation Finally, the authors consider the relevance
of small-world models for understanding the emergence of complex behaviors and the resilience of brain systems to pathological attack by disease or aberrant development.
They conclude that small-world models provide a powerful and versatile approach to understanding the structure and function of human brain systems.
![Page 5: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/5.jpg)
What Is a Small-World Network?
The basic idea of a small-world network is immediately familiar to many of us from personal experience.
We each have a social network of friends, relatives, and acquaintances.
Our close friends and relatives are likely to constitute a cluster or clique of social contacts; For example, two of my close friends are
likely to be friends with each other as well as with me.
![Page 6: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/6.jpg)
What Is a Small-World Network?
However, we may also have had the apparently surprising experience of traveling in a distant country or working in a strange city and discovering that some of the new people we meet in such a remote location are socially connected to people we already knew (friends of friends).
It appears that social networks, as well as being locally clustered or cliquish, are remarkably extensive:
The number of personal friendships mediating a social connection between any two people can be surprisingly small given the size and geographical dispersion of the global population.
![Page 7: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/7.jpg)
What Is a Small-World Network?
These qualitatively familiar concepts were translated to a more quantitative physical basis in an influential article by Watts and Strogatz (1998).
They constructed a simple computer model of a regular network or lattice, in which each node of the network was connected by a line or edge to each of its four nearest neighbors.
![Page 8: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/8.jpg)
What Is a Small-World Network?
This network structure or topology is highly clustered or cliquish by design, but to get from one node to another node on the opposite side of the lattice, one must traverse a large number of short-range connections.
In other words, although the path length (or number of mediating edges) between neighboring nodes is short, the path length between distant nodes is long, and so the minimum path length averaged over all possible pairs of nodes in the network is also long.
![Page 9: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/9.jpg)
What Is a Small-World Network?
Watts and Strogatz investigated the change in network topology (measured in terms of local clustering and minimum path length) that resulted from randomly rewiring some of the lattice edges to create long-range connections between distant nodes.
If many lattice edges were randomly rewired, the network naturally assumed the topological characteristics of a random graph (short path length and low clustering).
![Page 10: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/10.jpg)
What Is a Small-World Network?
But importantly, they found that the existence of even a few long-range connections greatly reduced the minimum path length of the network without affecting its local clustering to the same extent.
Thus, they defined algorithmically for the first time a class of networks with topological properties similar to social networks, demonstrating both the high clustering of a lattice and the short path length of a random graph, which they called small-world networks.
![Page 11: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/11.jpg)
What Is a Small-World Network?
![Page 12: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/12.jpg)
Regular
The computational model of small-world networks proposed by Watts and Strogatz (1998) began by connecting nodes with their nearest neighbors, producing a regular graph that had a high clustering coefficient and a high average path length.
![Page 13: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/13.jpg)
Random
With a probability P, edges were then randomly rewired. When P was equal to unity, all edges were randomly rewired, and thus the network was perfectlyrandom, having a short average path length and clustering coefficient.
![Page 14: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/14.jpg)
Small-World
However, when P was between 0 and 1, there existed some dense local clustering, characteristic of regular networks, and some long-range connections, characteristic of random networks (i.e., the resultant graph was a small-world network with high clustering and low path length).
![Page 15: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/15.jpg)
What Is a Small-World Network?
In the 8 years since small-world networks were described quantitatively in this way, there has been a remarkable profusion of studies seeking to clarify their mathematical properties and/or to explore their suitability as models of real-life networks.
For example, the distinctive combination of high clustering and short path length has been reported in:
Social networks of professionally collaborating actors or scientists. In infrastructural networks such as the Internet, power supply
grids, or transport systems. And in biochemical systems such as cellular networks of protein-
protein or gene-gene interactions. Small-world topology has been demonstrated empirically in
complex networks at physical scales ranging from molecular to macroeconomic and in scientific contexts as diverse as ecology, computing, and linguistics.
![Page 16: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/16.jpg)
Why Should We Think about the Brain as aSmall-World Network?
There are empirical and theoretical reasons a priori why small worlds present an attractive model for brain network connectivity. Later, we will review the
mathematical methodology and empirical findings in more detail.
first, we briefly rehearse the main theoretical motivations.
![Page 17: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/17.jpg)
Why Should We Think about the Brain as aSmall-World Network?
The brain is a complex network on multiple spatial and time scales. This fact alone might motivate a
small-world analysis of brain networks given the widespread occurrence of small-world properties in so many other complex networks and over a wide range of physical scales.
![Page 18: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/18.jpg)
Why Should We Think about the Brain as aSmall-World Network?
The brain supports both segregated and distributed information processing. Network architecture is regarded as a key
substrate for sensorimotor and cognitive processing, which may be localized discretely in specialized regions or represented by coherent oscillations in large-scale distributed systems.
Smallworld topology comprises both high clustering (compatible with segregated or modular processing) and short path length (compatible with distributed or integrated processing).
![Page 19: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/19.jpg)
Why Should We Think about the Brain as aSmall-World Network?
The brain likely evolved to maximize efficiency and/or minimize the costs of information processing.
Small-world topology is associated with high global and local efficiency of parallel information processing, sparse connectivity between nodes, and low wiring costs.
Small-world networks can operate dynamically in a critical state, facilitating rapid adaptive reconfiguration of neuronal assemblies in support of changing cognitive states.
![Page 20: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/20.jpg)
Mathematical Concepts
Complex systems can be better understood when we describe them mathematically as graphs. For graph analysis, the N individual
components or agents comprising the system are called nodes.
The K relations or connections between them are called edges.
![Page 21: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/21.jpg)
Mathematical Concepts
The edges of a graph can be directed or undirected: An undirected graph simply summarizes
symmetric relations (such as correlations) between nodes.
whereas a directed graph additionally models the causal relationships between nodes.
![Page 22: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/22.jpg)
Mathematical Concepts
Edges can also be categorized as weighted or unweighted: In an unweighted graph, all the edges
are assumed to indicate relations of equivalentstrength between nodes.
whereas a weighted graph can be used to differentiate stronger and weaker connections.
![Page 23: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/23.jpg)
Mathematical Concepts Each node of a graph can be described in
terms of the number of edges that connect to it: This is called the degree of a node, k. The nearest neighbors of a node are directly
connected to it by a single edge. In a directed graph, we can distinguish
between the in-degree (number of afferent edges) and the out-degree (number of efferent edges) of a node.
![Page 24: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/24.jpg)
Mathematical Concepts
The degree distribution of a graph is the probability distribution of k.
Random graphs have an exponential degree distribution: (Albert and Barabási 2002).
Several complex systems, such as the Internet and World Wide Web (WWW), have been found to have a power law distribution of the form , which implies a greater probability that nodes with a very large degree will exist in the graph
![Page 25: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/25.jpg)
Mathematical Concepts the power law degree distribution of the
WWW is compatible with the existence of a few major hubs such as Google or Yahoo!, to which very many other sites are linked.
Many physically embedded networks, such as transport or infrastructural systems, have an exponentially truncated power law distribution of the form ,which implies that the probability of highly connected hubs will be greater than in a random graph but smaller than in a scale-free network with a power law degree distribution.
![Page 26: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/26.jpg)
Mathematical Concepts
![Page 27: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/27.jpg)
Path Length
A, The path length between the two yellow nodes is defined as the fewest number of edges that must be traversed to get from one to the other. In this case, five edges must be followed, and therefore the path length between these two nodes is five.
![Page 28: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/28.jpg)
Clustering
B, A high clustering coefficient means that if two nodes are both connected to a third node, then they are probably also connected to each other. The calculation of the clustering coefficient takes into account the number of connected triangles (shown here with yellow nodes and dashed edges).
![Page 29: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/29.jpg)
Average Degree
C, The degree of a node is equal to the number of edges connected to it. A hub is defined as a node that has a degree larger than the average degree. The average degree in this network is 3.3, and therefore, both nodes with degree 6 are hubs (shown in yellow).
![Page 30: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/30.jpg)
The Two Key Metrics Of Small Worldness
The two key metrics of small worldness introduced by Watts and Strogatz (1998) are: The clustering coefficient C. The minimum path length L.
![Page 31: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/31.jpg)
Minimum Path Length L
The path length between any two nodes is simply defined as the minimal number of edges that must be traversed to form a direct connection between the two nodes of interest.
![Page 32: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/32.jpg)
Clustering Coefficient C
The clustering coefficient of a node is a measure of the number of edges that exist between its nearest neighbors.
If all the nearest neighbors of an index node are also nearest neighbors of each other, then C will have its maximum normalized value (1).
![Page 33: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/33.jpg)
The Two Key Metrics Of Small Worldness
The minimum path length and clustering coefficient of the whole graph are estimated simply by averaging Li and Ci for each of the i = 1, 2, 3, . . . N nodes it comprises.
To evaluate the properties of a real-life network, these parameters must be compared to the mean clustering coefficient and path length estimated in a random graph with the same number of nodes, edges, and degree distribution as the network of interest.
![Page 34: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/34.jpg)
The Two Key Metrics Of Small Worldness
In a random graph, the average minimum path length is typically short, Lrand ∼ ln N/ln([K/N] – 1), and the average clustering coefficient is typically small, Crand ∼ (K/N)/N (Albert and Barabási 2002).
For a small-world network, by definition, we expect the ratio λ = L/Lrand to be approximately 1 and the ratio γ = C/Crand to be greater than 1.
Therefore, a simple scalar measure of small worldness can be defined as σ = γ/λ, which will be greater than 1 if the network has the characteristic property of greater-than-random clustering and near-random path length (Humphries and others 2006).
![Page 35: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/35.jpg)
Alternative Metrics Of Complex Networks
More recent methodological work has developed alternative metrics of complex networks that can be related to C and L and may have some technical or conceptual advantages.
For example, we can define the global efficiency of information transmission by a network as inversely proportional to the average minimum path length, Eglob ∼ 1/L (Latora and Marchiori 2001).
![Page 36: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/36.jpg)
Global Efficiency Unlike path length, global efficiency can be
measured in networks that are not composed of a single, large group of interconnected nodes.
Global efficiency may also be preferable to path length as a metric of brain network topology because it is more immediately related to the functional efficiency of the system for information transmission between any two nodes via multiple parallel paths.
This is attractive, compared to the more serial metric of path length, because the brain is known to instantiate parallel processing.
![Page 37: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/37.jpg)
Local Efficiency We can also define the local efficiency or
fault tolerance of a network as proportional to the clustering coefficient, Eloc ∼ C.
Both these efficiency measures can be compared to the maximum global and local efficiency of an ideal network, in which all possible connections between nodes are present and thus normalized to the range
0 < (Eglob, Eloc) < 1.
![Page 38: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/38.jpg)
Cost Another potentially important metric for
brain network analysis is some measure of cost.
One simple cost function is the sum of edges between regions in the graph (Latora and Marchiori 2003). a costly network will have many edges.
Many small-world networks in biology and sociology have been shown to have the economic property of delivering high global and local efficiency at relatively low cost (Latora and Marchiori 2003).
![Page 39: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/39.jpg)
Brain Anatomical Networks
Are Small-World Networks more than a model ?
![Page 40: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/40.jpg)
First S-W Nervous System Caenorhabditis
elegans 282 Neurons 2462 Synaptic
Connections Neither random
nor regular L=2.65, C=0.28
![Page 41: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/41.jpg)
Brain Structural & Functional Connectivity Analysis Preceded the Small-World model
(early 1990s), but identified its features
Identified Hierarchies, streams, areas, functional segregation & integration, clusters, short paths, complexity
Attempted modeling using regular graphs until Small-World networks discovery
![Page 42: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/42.jpg)
NMDS (Non-metric Multidimensional Scaling)
![Page 43: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/43.jpg)
Modeling as Regular Graphs Nearest Neighbor
Only Neighboring areas are connected (0/1) Nearest Neighbor or Next Door But one
Neighboring areas: strong connections (weight: 3)
More than one nearest neighbor in common: moderate connections (weight 2)
Areas sharing one nearest neighbor: weak connections (weight 1).
![Page 44: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/44.jpg)
Nearest Neighbor ~20% of real
connections predicted
~31% false connections predicted
![Page 45: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/45.jpg)
Nearest Neighbor or Next Door But one ~48% of real
connections predicted
~49% false connections predicted
![Page 46: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/46.jpg)
Brain Anatomical Networks are Small-World
Small-World model discovered in 1998 (Watts and Strogatz)
Existing data of mammal cortical anatomical networks analyzed and shown to be Small-World in 2000 (Stephan and others, Hilgetag and others)
![Page 47: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/47.jpg)
If you torture the data enough, it will admit anything
![Page 48: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/48.jpg)
Why Evolution favors Small-World networks - intuition
Minimize Cost (wiring) Maximize Adaptivity
no need to relocate areas Maximize Complexity Optimize the economic problem of
cost-effective information processing
![Page 49: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/49.jpg)
Minimize Wiring Cost Selection Pressure – Physical Evidence Segregation of white
and gray matter Separation of visual
cortical areas Scaling of the number
of areas/neuronal density with brain size
Organization of cortical areas and basal ganglia
Existence of topographic maps
Ocular dominance patterns
Dimensions of axonal and dendritic arbors
The fraction of gray matter occupied by axons and dendrites
![Page 50: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/50.jpg)
The Problem with Minimum Wiring
Complete minimization would allow only local connections
No long distance connections Delayed information transfer Metabolic energy depletion
Evolution also pressures to minimize energy costs
![Page 51: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/51.jpg)
Dynamical Complexity and Computational S-W Networks
Ok, anatomically brain networks have Small-World properties
Where’s the computational gain ? Analyze dynamic properties: entropy,
integration and complexity Analyze synchronizability
![Page 52: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/52.jpg)
Functional Connectivity
Dates before S-W model Analyze networks as dynamical
systems Measures temporal correlation
between neuronal units (cells, areas) No regard to anatomy (units can be
anatomically remote) Model-free (statistical, not causal)
![Page 53: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/53.jpg)
Ancient Flame War: Localizationists vs. Holists
Evidence shows: the brain has specialized areas for specific functions
Evidence also shows: high level perception is an integration process done by “the brain”, not by its parts
Proposed solution: both capabilities are necessary: Introducing “complexity”
![Page 54: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/54.jpg)
Gaussian Multidimensional Stationary Stochastic Process Stochastic Process: to describe
dynamical system as change in random variable
Multidimensional: each node is a random variable
Gaussian: Entropy computed from COV Stationary: statistical properties fixed
over time (Linear: COV matrix can be computed
analytically from connection matrix)
![Page 55: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/55.jpg)
Functional Connectivity Measures Entropy – degree of statistical
independence Integration – degree of deviation
from statistical independence Complexity – degree of being both
small scale segregated (high entropy) and large scale coherent (high integration)
![Page 56: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/56.jpg)
Entropy Randomness
Fair coin has maximal entropy Completely unfair coin has no entropy
Self information (Surprise) What information do you gain from
knowing event x has occurred? How surprised are you?
Solely dependant on probability
![Page 57: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/57.jpg)
Integration
Given random variables X1 … Xn composing system X Σi=1..nH(Xi) - H(X) The difference between the combined
entropy of the components and the system’s entropy as a whole
The difference indicates system has an integration mechanism to lower entropy
![Page 58: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/58.jpg)
Complexity
Calculate whole system integration I(X)
Break system down to subsets in all sizes
Calculate the difference between average measured integration and expected integration (k/n * I(X))
Sum it all up
![Page 59: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/59.jpg)
Functional Connectivity Results
![Page 60: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/60.jpg)
Explaining the results
To achieve integration, network needs to be highly connected (little integration in case a)
But that’s not enough for high complexity Too much integration leads to global
coherence, and the whole network oscillates together (case c)
![Page 61: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/61.jpg)
Functional Connectivity Conclusions
High complexity related to highly clustered, yet having few long range connections (leading to short paths)
Clustering or average short paths alone -> low complexity
Sounds familiar ?
![Page 62: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/62.jpg)
Functional Connectivity and Small-World networks
=> Take an actual Small-World anatomical network with, ‘run’ as a dynamical system, extract COV matrix, compute E, I, C
<= Graph Selection (genetic algorithm): start with random graph; rewire randomly; select variant with maximal C; repeat
![Page 63: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/63.jpg)
Growing Small-World Networks
![Page 64: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/64.jpg)
Functional Connectivity Supports S-W hypothesis Grown graphs with high complexity
are Small-World Actual Small-World cortical
networks show high complexity High complexity = Low wiring
costs (close to minimum) High entropy/integration = High
wiring costs (close to random)
![Page 65: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/65.jpg)
Synchronizability
Synchronization thought to be crucial for information processing by the brain Synchronization may form dynamical
cell assemblies that bind features such as visual objects
Synchronization provides robustness against noise
Measure: how fast and how well?
![Page 66: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/66.jpg)
Synchronization in non Small-World Networks Random/all-to-all networks blur the
information, not allowing synchronization (and does not match anatomy – too many connections for actual neurons)
Locally clustered networks very slow on passing the information (and does not match anatomy – too long pair-wise distance)
Model as network of physical oscilators
![Page 67: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/67.jpg)
Fast Response and Coherent Oscillations
Random graphs = fast response, no coherent oscillations
Regular graphs = slow response, coherent oscillations
S-W graphs = fast response, coherent oscillations
![Page 68: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/68.jpg)
One picture is worth 1000 words Regular network:
slow response, weak oscillation, weak coherence
S-W network: fast response, strong oscillation, strong coherence
Random network: fast response, no oscillation
![Page 69: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/69.jpg)
Oscillation Amplitude & Degree of Coherence Upper graph is the
familiar clustering/path-length picture
Middle graph is oscillation amplitude
Lower graph is degree of coherence (wave form)
![Page 70: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/70.jpg)
Results S-W networks
overcome the slow response problem
Surprisingly, very quickly and using very few random connections
Highly dependant on P; K (coupling strength) saturates
![Page 71: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/71.jpg)
Small-World or Scale-Free?
Scale free is a competing model Also common in real world (WWW) Connections based on power law Speaks of special hub nodes (Google) Sensitive to hub tear-down S-W people say physiologically
unreasonable, no evidence (sampling ?)
![Page 72: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/72.jpg)
From Functional Connectivity to Undirected Graph Anatomical data is ‘analog’ Computational models are ‘digital’ Applying computational tools on
anatomical data requires conversion: sampling & thresholding
Questions are: What do you sample ? How do you filter ?
![Page 73: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/73.jpg)
Problem
Reported topology of brain functional networks highly depends on: The number of regions included The chosen measure of association The thresholding rule
Example: Scale-Free model (Rare) hubs easily missed during
sampling
![Page 74: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/74.jpg)
Proof ?
![Page 75: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/75.jpg)
Small World analysis with FMRI
FMRI-functional Magnetic Resonance Imaging.
narrow bandwidth~0.01-1.0 Hz (comparing to 0.01-100 Hz in EEG\MEG).
This is limitation for try understanding small world brain functionalNetworks, however this tool can still be useful since it has been shown (Salvador, Suckling, Coleman, and others 2005) that long range connections (between frontal and parietal cortex) are stronger at frequencies less that 0.1 HZ)
![Page 76: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/76.jpg)
![Page 77: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/77.jpg)
Mathematical tools for fMRI analysis
We need mathematical tools for the estimation of functional connectivity between fMRI time series (converts signal from time domain into the frequency domain)
Fourier Transform
Wavelet Transform
![Page 78: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/78.jpg)
fMRI first demonstration test on humans brain functional (Salvador, Suckling, Coleman and others 2005)
fMRI time series were recorded from five healthy people in 90
Cortical and sub cortical regions in resting state.
The cortical region consists of the cerebral cortex it’s main fields are memory, attention, thought, language, and consciousness.
The sub cortical consists of the thalamus, hypothalamus, cerebellum (little brain) and brain stem.
![Page 79: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/79.jpg)
![Page 80: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/80.jpg)
fMRI first demonstration cont’
Functional connectivity was estimated between each possible pair of time series, using the Fourier transform and the partial correlation coefficient.
The connectivity matrix built with randomly threshold.
![Page 81: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/81.jpg)
fMRI first demonstration cont’ 2
Multivariate analysis of the functional connectivity matrix was performed using Hierarchical Clustering and metric multidimensional scaling.
It has shown many similarities to anatomical matrix that previously described by Multivariate analysis, like dense local connections between regions comprising visual, somatosensorimotor and auditory-verbal clusters, and sparse connectivity between functionally segragated dorsal and ventral components of the visual cortex.
![Page 82: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/82.jpg)
Wavelet based functional connectivity analysis(Achard and others, 2006)
As we have mentioned, Wavelet transform is mathematical alternative to Fourier transform.
Wavelet based functional connectivity analysis was introduced by Achard and colleagues, who reported small world properties of fMRI,networks at all frequencies in the range 0.007-0.45Hz, when the most significant was in the frequency interval 0.03-0.06Hz.
![Page 83: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/83.jpg)
Wavelet based cont’
They have shown that at a regional level of analysis, clustering was negatively correlated with the physical distance of connections to a node.
This result also have been made in analysis of anatomical connectivity matrices, and proposed a new way to classify cortical areas either as highly specialized (locally clustered connections), or integrative (long distance connections; Spons and Zwi 2004).
![Page 84: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/84.jpg)
![Page 85: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/85.jpg)
Network degree distribution
The degree distribution of a complex network (natural network) potentially provides an important clue to the processes constraining it’s formation and growth.
An example for that is the power law degree distribution of the WWW is compatible with it’s growth by creation of
new nodes, which preferentially form connections to existing hubs (like Google, yahoo, or any other big servers).
The whole brain fMRI networks of cortical regions has also been described as an exponentially truncated power law.
![Page 86: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/86.jpg)
Some facts…
Existence of Tpld in brain functional networks could likewisereflect: 1) the formative constrains of age of regions (Dorogovtsev & Mendes 2000).
2) metabolic cost of maintaining long range connections (Xulvi-Brunet & Sokolov 2002), or an upper limit on the number of connections that a single region can accommodate (Albert & Barabasi 2000).
![Page 87: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/87.jpg)
Facts cont’
3) small world networks can be created in a multitude of ways:
a) randomly rewiring links of a regular graph (Walts and Strogatz 1998).
b) adding new links to it with probability P (Newman and Watts 1999).
c) let sites connect locally to geographically nearby sites (Ozik & others 2004). and many more (see the paper for all of them..)important question in this context for future research is which of these growth rules are most important biologically.
![Page 88: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/88.jpg)
Emergent Properties and Pathology of Small-World Networks
The topology of a brain network dictates not only it’s possible physical dynamics, but also it’s emergent properties.
It has been shown (Simard & others 2005) that small-world network allow higher rates of information processing and learning than random graphs do.
One way to understand it is that changes in cognitive state (or cognitive capacity) might be connected with changes (potentially rapid) in the configuration of brain functional networks.
![Page 89: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/89.jpg)
Emergent Properties cont’
Nowadays, not a lot is known about the cognitive correlates in small world parameters of fMRI or EMG/EEG networks, though in an experiment held by Stam & colleagues (2006) they reported that cognitive decline due to Alzheimer disease is associated with increase path length (or reduced global efficiency) of EEG networks in the betha band.
![Page 90: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/90.jpg)
Functional resilience of brain networks
Functional resilience of brain networks to pathological attack can be modeled by deleting one or more nodes from the network and re estimating it’s small world parameters.
In deletion, if we delete one node (doesn’t matter which one) than we expect to increase path length (reduce global efficiency), but deletion of Hubs could have especially dramatic effect on overall network.
![Page 91: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/91.jpg)
Functional resilience cont’
Measurements of the path length of a brain functional network which was degraded by random deletion of nodes and by targeted attack on the association cortical hubs showed (Achard & colleagues 2006) that the brain network was resilient to random attack (and especially to targeted attack) more than comparable graph with a power law degree distribution.
![Page 92: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/92.jpg)
Functional resilience cont’
This fact can tell us that not only the Tpld is possibly reflecting physical constrains on network growth, but it could also reflect a selection of pressure favoring network topologies that are functionally resistant to local pathological attack.
It’s likely possible that understanding this idea could help in the future to know more about some brain deceases like epileptic seizures.
![Page 93: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/93.jpg)
![Page 94: Small-World Brain Networks Danielle Smith Bassett and Ed Bullmore Written By: Gassan Tabajah Ron Adar Gil Rapaport.](https://reader035.fdocuments.in/reader035/viewer/2022070323/56649d395503460f94a1392c/html5/thumbnails/94.jpg)
Conclusion
Nervous systems are anatomically connected as small world networks at macro and micro scales.
In humans, there have been recent reports of small world brain functional networks measured using fMRI or MEG\EEG, but there is still a lot to learn about the (presumably small world) topology of human brain anatomical networks, their development, emergent properties, and impact of neuropsychiatric disorders and drugs on small world brain networks.