From Connection to Cognition
Yudith Haveman
1 Introduction
The word cognition originates from cognitio in Latin, meaning knowledge 1. Although there is not
one clear definition of cognition, the most influential one is of Neisser (1967) 2, referring to cognition
as the mental process by which external or internal input is transformed, reduced, elaborated,
stored, recovered, and used. As such, it involves functions including attention, memory, decision
making, planning and executing actions 1. Research implicated that cognition seems to be impaired
in a variety of conditions, such as in neurological and psychiatric disorders, and even the natural
process of aging 3–7. For example, Alzheimer’s Disease (AD) is predominantly characterized by
impaired short term memory. With disease progression, disturbances of other memory functions,
language ability, praxis, visuospatial and executive functions are seen as well 7. Neurologically,
this was related to gray matter loss in the posterior cingulate cortex, hippocampus, precuneus, insula
and dorsolateral prefrontal cortex 8. In addition, the activity of the medial temporal lobe was
decreased, while the ventral lateral prefrontal cortex was more activated in AD 9, and there were
alterations seen in the default mode network, fronto-parietal and salience network 8. This indicates
that cognitive impairment relies on disturbed brain regions spanning multiple lobes. So, we can
split up cognition into different processes, but mapping cognition directly onto the cortex is more
complex. Therefore, a promising paradigm is that cognition can be best approached as brain areas
working together as large-scale networks (i.e., neural networks distributed across many areas)10–13.
1
2 Measuring Large-scale brain networks
Large-scale brain networks consist of a collection of brain regions (nodes) and their connections
(edges). These two characteristics can be measured at a structural level, in which the nodes represent
the gray and white matter regions in the brain 10–12. This can be measured using structural
magnetic resonance imaging (sMRI) using a threshold for cortical thickness or an automated
anatomical labeling (AAL) template 14 to parcellate the whole cerebral cortex into different areas.
At a structural level, these nodes are linked with white matter tracts (i.e., edges). These white
matter tracts can be measured using diffusion tensor imaging (DTI) 10–12. A dysfunctional node
and altered white matter tracts can bi-directionally affect each other.
Networks can also be constructed via functional connectivity. 10–12, 15. With this approach, simultaneously
active brain regions during cognitive tasks are considered functionally connected 15.
Functional connectivity measurements are based on blood oxygenation level–dependent (BOLD)
functional MRI (fMRI) or coherence in electro- or magnetoencephalogram (EEG/MEG) signals
acquired during task performance or resting state. Important is that the identification of edges depends
on the monitoring methodology 10; whether edges are identified in the time-domain (crosscorrelation)
or frequency (spectral coherence or phase synchronicity) domain; whether to use independent
component analysis or seedbased functional interpedence analysis. In the latter method,
a seed region associated with a cognitive function is identified, after which a map is constructed
with simultaneously active brain areas. Another technique to determine functional connectivity is
positron emission tomography (PET), in which connectivity is based on glucose metabolism 10, 11.
2
With the use of these different techniques to identify nodes and edges, a graph-theoretical
analysis of brain networks can be computed, which can be compared between different conditions
and cognitive abilities.
3 Graph Theory and Analysis
A graph theory is a mathematical representation that decomposes networks into nodes and edges
11, 15. Main measures to describe graphs are (a) degree, i.e., the number of connections from the
node of interest to other nodes of the network, (b) path length, i.e., the minimum number of edges
traversed to go from one node of interest to the other, (c) efficiency, which is inversely related
to the path length, (d) clustering coefficient, which refers to the number of connections between
the nearest neighbours of a node as a proportion of the maximum number of connections, and (e)
betweenness centrality, which measures how many of the shortest paths between all other node
pairs in the network pass through this node, indicating the importance of a node 11, 15.
Based on these measurements, three different network architectures are found, which lie on
spectra regarding clustering coefficients and path lengths 12, 15, 16. A regular network has on one side
a high clustering coefficient, but on the other side long path lengths. Random networks are highly
disordered and are exactly the opposite of regular networks, with low clustering coefficient and
short path lengths. In between lie small-world networks, characterized by a relatively high degree
of order, but also a number of long-range connections, resulting in networks with high clustering
and low path lengths. It was shown that a high clustering coefficient was beneficial, because it is
3
associated with high local efficiency of information transfer and robustness 12, 15–18. In addition, a
short path length was advantageous as well, because it indicates a high global efficiency of parallel
information transfer. Therefore small-world networks in the brain are optimal, switching to either
regular or random networks can cause impairments, for example in cognition 12, 15–18.
4 Graph Theory in Cognition
Studies have shown that the small-world architecture is related to cognitive performance. For
example, Li et al. (2009)19 showed that intelligence quotient scores are positively correlated with
small-world network properties. During aging, a decline of cognition is seen, accompanied with
longer path lengths, increased clustering coefficient, decreased cortical connectivity and decreased
efficiency 20–23. So, the network properties were switched from a small-network architecture to a
more regular network, perhaps because of a reduced ability to integrate information on a global
level. Other studies showed that nodes in the frontal cortex, insula and hippocampus got a lower
betweenness centrality during aging, indicating that these nodes became less important 22, 23. In
short, disturbances in network architecture as consequence of aging lead to less optimal network
properties, in which the regions prominently important for cognition are most affected. Given the
impact of normal aging on brain networks and correlations with cognition, one can imagine that
changes in network properties in diseased states, such as AD, can result in impaired cognition.
4
5 Disturbed brain networks in Alzheimer’s Disease
Delbeuck et al. (2003)24 proposed that AD is a dysconnection syndrome in which damage of amyloid
beta at important nodes results in brain network impairments and finally cognitive dysfunction.
There are less damaged nodes seen in mild cognitive impairment (MCI)25, suggesting that changes
in the network may reflect changes in cognition.
At a structural level, Lo et al. (2010)26 used DTI to construct white matter networks and
correlate graph properties to cognition in AD. Nodes were defined with the AAL atlas including
78 brain regions, edges were generated based on DTI data. The AD patients exhibited longer path
lengths, lower global efficiency, and reduced local efficiency, mainly frontally. Path length and
global efficiency were related to verbal memory 26. This study showed that a less optimal network
architecture in AD may result in cognitive impairment. Other studies used sMRI in which nodes
were defined based on cortical thickness and their correlation matrices were used to define edges
27–29. Some studies showed that path length and clustering coefficient were again higher in AD
patients compared to healthy controls, which suggests to a more regular network 27, 30. In addition,
MCI patients scored on these graph measurements in between healthy controls and AD patients 30.
Contrary, studies found that the clustering coefficient and mean path length were reduced in the
MCI and AD groups 28, 31. Regression analysis showed that the mean path length correlated with
MMSE scores 31. These contrary results might be due to the choice of correlation matrix (pearson
or partial) 29. Still, it can be said that a loss of small-world characteristics in AD is related to
cognitive impairment.
5
At a functional level, the loss of small-world features in AD is seen is well, but the direction
of these changes is contradictory. For example, Stam et al. (2009) 32 conducted an MEG study with
149 channels as nodes and edges were determined by synchronization between the channels using
a phase lag index. Results showed a lower clustering coefficient and path length in the alpha band,
indicating a more random network in AD patients 32, 33. The extent of network randomization
correlated with cognition in AD. In addition, there was seen a decreased synchronization in the
alpha and beta bands, which was predominantly present in the important nodes 32. Other MEG
studies 34 and EEG studies 33, 35 showed that an increase in low-frequency bands and decrease in
alpha frequencies were associated with a disconnected network in AD. However, Stam et al. (2007)
36 found a longer path length in AD patients, which correlated with cognition. Here, the nodes were
based on 21 EEG channels, edges were determined with synchronization likelihood of the nodes.
These contradictory findings are also seen in the fMRI studies: some studies found a lowering
clustering coefficient and path length, and loss of long-distance connections in AD 37, 38. PET
studies added that the connectivity was especially reduced in the prefrontal cortex and important
nodes showed greater hypometabolism and reduced connectivity 39, 40. On the other side, some
studies found an increased clustering coefficient, path length, local efficiency and betweenness
centrality in AD patients 41, 42. Independent of the direction of the changing network features, the
small-world networks do change in correlation with cognitive impairment.
With the use of different neuroimaging techniques to define structural and functional networks,
the relation between cognition and small-network features is present. However, the use
of different imaging techniques results in different directions of network features with impaired
6
cognition. First of all, structural and functional connectivity differ. For example, structural connectivity
can be decreased in neurodegenerative disorders, but at a functional level compensation
is possible, resulting in increased functional connectivity. Second, when only looking to functional
connectivity, comparing different imaging techniques can also be problematic since MEG
and EEG measure only cortical changes in activation, whilst fMRI can assess subcortical network
changes. Differences in resolution can also influence the graph measurements. Third, even the
results obtained by the same techniques can differ, due to choices in the use of particular statistical
correlations 31, definition of nodes and edges, spatial smoothing problems with atrophy and the
composition of the sample group.
6 Conclusion
Graph analysis is a powerful tool to construct the network architecture of a brain. A small-world
network is the most efficient network in both local processing and global integration of information.
A variety of conditions, including AD, change this efficient network into a more regular or
random one, which may underlie cognitive dysfunction. While it has been shown that changes in
network properties occur with cognitive impairments, further study with graph analysis is needed
to elucidate fully the nature and effect of network changes.
References
1. Brandimonte, M., Bruno, N., Collina, S., Cognition, P. P. & d’Ydewalle, G. Psychological
concepts: an international historical perspective (psychology press, 2006).
7
2. Neisser, U. Cognitive Psychology New York: Appleton-Century-Crofts (Psychology Press,
1967).
3. Hofer, S. M. & Alwin, D. F. Handbook of cognitive aging: Interdisciplinary perspectives
(Sage, 2008).
4. Millan, M. J. et al. Cognitive dysfunction in psychiatric disorders: characteristics, causes and
the quest for improved therapy. Nature reviews Drug discovery 11, 141 (2012).
5. Medalia, A. & Lim, R. Treatment of cognitive dysfunction in psychiatric disorders. Journal
of Psychiatric Practice 10, 17–25 (2004).
6. Langdon, D. W. Cognition in multiple sclerosis. Current opinion in neurology 24, 244–249
(2011).
7. Kester, M. I. & Scheltens, P. Dementia: the bare essentials. Practical neurology 9, 241–251
(2009).
8. Kinkingnehun, S. et al. Vbm anticipates the rate of progression of alzheimer disease: a 3-year
longitudinal study. Neurology 70, 2201–2211 (2008).
9. Schwindt, G. C. & Black, S. E. Functional imaging studies of episodic memory in alzheimer’s
disease: a quantitative meta-analysis. Neuroimage 45, 181–190 (2009).
10. Bressler, S. L. & Menon, V. Large-scale brain networks in cognition: emerging methods and
principles. Trends in cognitive sciences 14, 277–290 (2010).
8
11. Bullmore, E. & Sporns, O. Complex brain networks: graph theoretical analysis of structural
and functional systems. Nature Reviews Neuroscience 10, 186 (2009).
12. Menon, V. Large-scale brain networks and psychopathology: a unifying triple network model.
Trends in cognitive sciences 15, 483–506 (2011).
13. Park, H.-J. & Friston, K. Structural and functional brain networks: from connections to cognition.
Science 342, 1238411 (2013).
14. Tzourio-Mazoyer, N. et al. Automated anatomical labeling of activations in spm using a
macroscopic anatomical parcellation of the mni mri single-subject brain. Neuroimage 15,
273–289 (2002).
15. Gits, H. C. Relating connectivity and graph analysis to cognitive function in alzheimer’s
disease. Michigan Journal of Medicine 1 (2016).
16. Watts, D. J. & Strogatz, S. H. Collective dynamics of ‘small-world’networks. nature 393, 440
(1998).
17. Bassett, D. S. & Bullmore, E. Small-world brain networks. The neuroscientist 12, 512–523
(2006).
18. Watts, D. J. Networks, dynamics, and the small-world phenomenon. American Journal of
sociology 105, 493–527 (1999).
19. Li, Y. et al. Brain anatomical network and intelligence. PLoS computational biology 5, 1–17
(2009).
9
20. Wu, K. et al. Age-related changes in topological organization of structural brain networks in
healthy individuals. Human brain mapping 33, 552–568 (2012).
21. Gong, G. et al. Age-and gender-related differences in the cortical anatomical network. Journal
of Neuroscience 29, 15684–15693 (2009).
22. Wang, L., Li, Y., Metzak, P., He, Y. & Woodward, T. S. Age-related changes in topological
patterns of large-scale brain functional networks during memory encoding and recognition.
Neuroimage 50, 862–872 (2010).
23. Zhu, W. et al. Changing topological patterns in normal aging using large-scale structural
networks. Neurobiology of aging 33, 899–913 (2012).
24. Delbeuck, X., Van der Linden, M. & Collette, F. Alzheimer’disease as a disconnection syndrome?
Neuropsychology review 13, 79–92 (2003).
25. Rombouts, S. A., Barkhof, F., Goekoop, R., Stam, C. J. & Scheltens, P. Altered resting state
networks in mild cognitive impairment and mild alzheimer’s disease: an fmri study. Human
brain mapping 26, 231–239 (2005).
26. Lo, C.-Y. et al. Diffusion tensor tractography reveals abnormal topological organization in
structural cortical networks in alzheimer’s disease. Journal of Neuroscience 30, 16876–16885
(2010).
27. He, Y., Chen, Z. & Evans, A. Structural insights into aberrant topological patterns of largescale
cortical networks in alzheimer’s disease. Journal of Neuroscience 28, 4756–4766 (2008).
10
28. Li, Y. et al. Discriminant analysis of longitudinal cortical thickness changes in alzheimer’s
disease using dynamic and network features. Neurobiology of aging 33, 427–e15 (2012).
29. Phillips, D. J. et al. Graph theoretic analysis of structural connectivity across the spectrum
of alzheimer’s disease: The importance of graph creation methods. NeuroImage: Clinical 7,
377–390 (2015).
30. Yao, Z. et al. Abnormal cortical networks in mild cognitive impairment and alzheimer’s disease.
PLoS computational biology 6, 1–11 (2010).
31. Tijms, B. M. et al. Single-subject grey matter graphs in alzheimer’s disease. PloS one 8, 1–9
(2013).
32. Stam, C. et al. Graph theoretical analysis of magnetoencephalographic functional connectivity
in alzheimer’s disease. Brain 132, 213–224 (2008).
33. de Haan, W. et al. Functional neural network analysis in frontotemporal dementia and
alzheimer’s disease using eeg and graph theory. BMC neuroscience 10, 1–12 (2009).
34. de Haan, W. et al. Disrupted modular brain dynamics reflect cognitive dysfunction in
alzheimer’s disease. Neuroimage 59, 3085–3093 (2012).
35. Vecchio, F. et al. Cortical brain connectivity evaluated by graph theory in dementia: a correlation
study between functional and structural data. Journal of Alzheimer’s Disease 45, 745–756
(2015).
11
36. Stam, C. J., Jones, B., Nolte, G., Breakspear, M. & Scheltens, P. Small-world networks and
functional connectivity in alzheimer’s disease. Cerebral cortex 17, 92–99 (2006).
37. Supekar, K., Menon, V., Rubin, D., Musen, M. & Greicius, M. D. Network analysis of intrinsic
functional brain connectivity in alzheimer’s disease. PLoS computational biology 4, 1–11
(2008).
38. Sanz-Arigita, E. J. et al. Loss of ‘small-world’networks in alzheimer’s disease: graph analysis
of fmri resting-state functional connectivity. PloS one 5, 1–14 (2010).
39. Seo, E. H. et al. Whole-brain functional networks in cognitively normal, mild cognitive impairment,
and alzheimer’s disease. PloS one 8, 1–11 (2013).
40. Sanabria-Diaz, G., Mart´ınez-Montes, E. & Melie-Garcia, L. Glucose metabolism during resting
state reveals abnormal brain networks organization in the alzheimer’s disease and mild
cognitive impairment. PloS one 8, 1–25 (2013).
41. Zhao, X. et al. Disrupted small-world brain networks in moderate alzheimer’s disease: a
resting-state fmri study. PloS one 7, 1–12 (2012).
42. Kim, H. et al. Non-monotonic reorganization of brain networks with alzheimer’s disease
progression. Frontiers in aging neuroscience 7, 111 (2015).
12