334
V. Vuksanovi´ c
independence. At the same time, detecting modular structures that underpin specific
function can be identified by characterizing interactions between those nodes that
show relatively similar activity/dynamics [29]. Likewise, meta-analysis on more than
1000 fMRI has shown the existence of functional modules specialized for specific
cognitive processes [21].
The brain appears to be divided into ’functional modules’ whose intra-modular
connectivity reflects the underlying structural (axonal) connections [38]. However,
although functional modules usually mirror local brain anatomy, they also incorporate
long-range interactions (i.e., those between spatially distant brain areas) [31, 75,
78]. More pertinent to this paper, the modular topology of brain functional (MRI)
networks is documented across different parcellations of the cortex (i.e., brain atlases)
[7, 49, 74]. Modularity as a property of morphology has been widely studied in the
context of evolution and development [48]. Recent neuroimaging studies suggest
modular organization of cortical morphology across regional thickness [74, 79],
surface area [61] or volume [3]. There is consistency in the organisation of these
networks whether they are based on correlating these features across individuals
within one group [61, 73, 74] or correlating regional features of an individual brain
[62]. The brain modular, yet integrated, functional organisation lowers the wiring
cost (i.e., the average length and number of connections) of the network [6], thus
potentially lowering metabolic costs [8] while providing more efficient information
processing [65]. More importantly, modularity, as mapped by large-scale brain fMRI
networks, is cognitively and behaviorally relevant; for example, it correlates with
variations in working memory [81] (Fig. 21.2).
21.2.4 Dynamical Functional Networks
An additional ‘layer’ to modular organisation of brain networks is the notion of
dynamical functional networks. In this context, the focus is on how likely regions are
to change their “module allegiance” and synchronize their activity with a different
set of nodes. The analysis of changes in network interactions over time utilises
non-stationary, time-varying dynamics of neuro-imaging recordings. Up to this
point, I have reviewed some of methods to map functional connections which predominantly utilize static network approaches (in which network edges remain constant
throughout time) derived from graph theory [14, 24]. However, such approaches
are unable to characterize or identify changes in regional interactions over time.
Furthermore, the emergence of dynamic functional networks from static structural
connections may resolve a fundamental understanding of how structure and function
map onto each other.
A promising way to obtain a fundamental understanding of how patterns of functional connectivity change over time is the simulation of brain dynamics using a
sophisticated modeling framework that implements nonlinear Kuramoto-like dynamics on a physical network backbone informed by both structural (white matter)
and functional (fMRI) connectivity maps [11, 76, 78]. At the same time, the
V. Vuksanovi´ c
independence. At the same time, detecting modular structures that underpin specific
function can be identified by characterizing interactions between those nodes that
show relatively similar activity/dynamics [29]. Likewise, meta-analysis on more than
1000 fMRI has shown the existence of functional modules specialized for specific
cognitive processes [21].
The brain appears to be divided into ’functional modules’ whose intra-modular
connectivity reflects the underlying structural (axonal) connections [38]. However,
although functional modules usually mirror local brain anatomy, they also incorporate
long-range interactions (i.e., those between spatially distant brain areas) [31, 75,
78]. More pertinent to this paper, the modular topology of brain functional (MRI)
networks is documented across different parcellations of the cortex (i.e., brain atlases)
[7, 49, 74]. Modularity as a property of morphology has been widely studied in the
context of evolution and development [48]. Recent neuroimaging studies suggest
modular organization of cortical morphology across regional thickness [74, 79],
surface area [61] or volume [3]. There is consistency in the organisation of these
networks whether they are based on correlating these features across individuals
within one group [61, 73, 74] or correlating regional features of an individual brain
[62]. The brain modular, yet integrated, functional organisation lowers the wiring
cost (i.e., the average length and number of connections) of the network [6], thus
potentially lowering metabolic costs [8] while providing more efficient information
processing [65]. More importantly, modularity, as mapped by large-scale brain fMRI
networks, is cognitively and behaviorally relevant; for example, it correlates with
variations in working memory [81] (Fig. 21.2).
21.2.4 Dynamical Functional Networks
An additional ‘layer’ to modular organisation of brain networks is the notion of
dynamical functional networks. In this context, the focus is on how likely regions are
to change their “module allegiance” and synchronize their activity with a different
set of nodes. The analysis of changes in network interactions over time utilises
non-stationary, time-varying dynamics of neuro-imaging recordings. Up to this
point, I have reviewed some of methods to map functional connections which predominantly utilize static network approaches (in which network edges remain constant
throughout time) derived from graph theory [14, 24]. However, such approaches
are unable to characterize or identify changes in regional interactions over time.
Furthermore, the emergence of dynamic functional networks from static structural
connections may resolve a fundamental understanding of how structure and function
map onto each other.
A promising way to obtain a fundamental understanding of how patterns of functional connectivity change over time is the simulation of brain dynamics using a
sophisticated modeling framework that implements nonlinear Kuramoto-like dynamics on a physical network backbone informed by both structural (white matter)
and functional (fMRI) connectivity maps [11, 76, 78]. At the same time, the