The Human Brain Network - Biological Networks - page 212

9.4. Network Topology and Network Disease
A number of recent studies of natural and technological networks have included
analyses that investigated the robustness or vulnerability of such network to
disruptions of their connection patterns. The scale-free network of the World
Wide Web has been shown to be robust with respect to random deletion of nodes,
but vulnerable to targeted attack on heavily connected hubs (Albert et al ., 2000;
Doyle et al ., 2005). Such targeted attacks on hubs often result in disintegration
or disconnection of the overall network. The fact that the robustness of social
and technological networks is linked to features of network topology raises the
interesting question if similar relations may be found for large-scale brain
networks.
In the brain, the mapping of functional deficits to underlying structural
perturbations is experimentally challenging, but essential for a more complete
understanding of brain damage and recovery. Structural perturbations may
include deletions of nodes or edges, or the rewiring of edges to new positions in
the network. It is currently unknown which structural measures best capture the
potential effects of node or edge lesions. A candidate measure of edge
vulnerability (Kaiser and Hilgetag, 2004) helps to identify edges whose loss most
affects global structural measures. It turns out that such edges often correspond
to “bridges”, i.e. edges linking segregated clusters of brain regions. More
recently, Kaiser et al . (2007) studied the effects of removing single areas on the
structural integrity of large-scale anatomical networks of the cat and macaque
cortex. One main finding indicates that lesions of “bottlenecks”, i.e. hubs that
link areas across segregated clusters, produce greater effects on structural
integrity than lesions of non-hub regions. The clustered architecture of the cortex
and the existence of a small number of connector hubs thus may result in greater
overall robustness of the network against randomly introduced lesions.
The dynamic consequences of lesioning large-scale brain networks have been
investigated in the context of lesion effects on the statistical structure of
spontaneous resting state activity (Honey and Sporns, 2008). To understand the
effects of a cortical lesion it is necessary to consider not only the loss of local
neural function, but also the lesion-induced changes in the larger network of
endogenous oscillatory interactions in the brain. To assess these nonlocal effects,
a computational model of the macaque visual and somatomotor cortex was
designed. The model demonstrated that lesions of single brain regions had
effects that extended beyond the immediate neighbors of the lesion site, and the
amplitude and dispersal of nonlocal effects are critically influenced by the way
areas of the network are organized into clusters. In the model, lesions of hubs