7 Computational EEG Analysis for the Diagnosis of Psychiatric …
151
changes of amplitude and latency of P3 have been confirmed in various neuropsychiatric disorders.
7.2.2 Functional Connectivity Network
Functional network analysis can be conducted on a wide range of EEG data. One
material that can be used universally to examine functional connectivity is the resting
state EEG. Resting state EEG holds many advantages because it is easy to obtain,
relatively stable, not difficult to handle, and independent from task characteristics.
Yet, several inevitable qualities require researchers to be cautious when dealing with
resting state EEG. For example, resting state EEG that is measured during an eyesclosed condition can contain unwanted artifacts, which result from drowsiness. This
leads to questions regarding the length and the number of epochs required in order to
obtain stable statistical power and how expert consensus on the standardization for
data qualification ought to be driven. While these points can pose some threat as to
the reliability of the measure, they are not insoluble and can be adequately addressed
[33]. Resting state EEG clearly has more advantages than weaknesses. Hence, EEG
functional connectivity of the resting state will be mainly discussed throughout this
text.
An increasing number of researchers have assumed that alterations in the cortical
connectivity network might provide insights to the underlying neural mechanisms of
mental illnesses. Many of these studies adopted the graph theory to quantify global
and local changes in the cortical functional connectivity network [12, 115, 121, 125].
In particular, the small-world network has been regarded as one of the most suitable
models to elucidate information transfer in the human brain [10]. The small-world
network is the middle ground between random network and regular network, and is
characterized by a higher clustering coefficient than random networks, and a shorter
path length than regular networks. The clustering coefficient and the path length each
reflect the amount of segregation of highly inter-connected units and the amount of
integration of the whole network [140]. Therefore, the small-world characteristics of
the brain allow for more efficient information transfer among distant brain regions.
On the other hand, a weighted network is a network where the ties among nodes
have weights assigned to them. Because of this nature, weighted networks are more
difficult to analyze than unweighted binary networks in which ties are simply present
or absent. Despite the difficulty of its analysis, a number of network measures have
been proposed for weighted networks including the following:
• Strength, which refers to the strength of the connection in the network, is estimated
by the sum of weights of links connected to the brain regions.
• Clustering coefficient (CC), which refers to the degree in which a node is clustered
with neighbor nodes, is calculated for the whole network.
• Path length (PL), which refers to the summation of lengths between two nodes in
the whole network, indicates overall connectedness of the whole network.
151
changes of amplitude and latency of P3 have been confirmed in various neuropsychiatric disorders.
7.2.2 Functional Connectivity Network
Functional network analysis can be conducted on a wide range of EEG data. One
material that can be used universally to examine functional connectivity is the resting
state EEG. Resting state EEG holds many advantages because it is easy to obtain,
relatively stable, not difficult to handle, and independent from task characteristics.
Yet, several inevitable qualities require researchers to be cautious when dealing with
resting state EEG. For example, resting state EEG that is measured during an eyesclosed condition can contain unwanted artifacts, which result from drowsiness. This
leads to questions regarding the length and the number of epochs required in order to
obtain stable statistical power and how expert consensus on the standardization for
data qualification ought to be driven. While these points can pose some threat as to
the reliability of the measure, they are not insoluble and can be adequately addressed
[33]. Resting state EEG clearly has more advantages than weaknesses. Hence, EEG
functional connectivity of the resting state will be mainly discussed throughout this
text.
An increasing number of researchers have assumed that alterations in the cortical
connectivity network might provide insights to the underlying neural mechanisms of
mental illnesses. Many of these studies adopted the graph theory to quantify global
and local changes in the cortical functional connectivity network [12, 115, 121, 125].
In particular, the small-world network has been regarded as one of the most suitable
models to elucidate information transfer in the human brain [10]. The small-world
network is the middle ground between random network and regular network, and is
characterized by a higher clustering coefficient than random networks, and a shorter
path length than regular networks. The clustering coefficient and the path length each
reflect the amount of segregation of highly inter-connected units and the amount of
integration of the whole network [140]. Therefore, the small-world characteristics of
the brain allow for more efficient information transfer among distant brain regions.
On the other hand, a weighted network is a network where the ties among nodes
have weights assigned to them. Because of this nature, weighted networks are more
difficult to analyze than unweighted binary networks in which ties are simply present
or absent. Despite the difficulty of its analysis, a number of network measures have
been proposed for weighted networks including the following:
• Strength, which refers to the strength of the connection in the network, is estimated
by the sum of weights of links connected to the brain regions.
• Clustering coefficient (CC), which refers to the degree in which a node is clustered
with neighbor nodes, is calculated for the whole network.
• Path length (PL), which refers to the summation of lengths between two nodes in
the whole network, indicates overall connectedness of the whole network.