226

Bioinformatics of the Brain

the difference in these values between two regions is used to determine the

edge weight of the edge between them to construct the overall graph.

• Effective Brain Networks (EBN): This type of connectivity shows the casual

interactions between brain regions by considering the influence of one brain

region to another.

We will focus on the construction of FBNs here and the analysis of these

networks throughout this review as this type of brain network is the subject of

various research studies to analyze brain functions. The main steps of building

such a network are as follows: [2]:

1.

Defining the nodes: Dividing the brain into large-scale homogeneous

and non-overlapping regions is performed to define the nodes of

the network. Selection of these regions is called parcellation which

is the process of dividing the brain into distinct regions based on

anatomical, morphological or topological criteria.

2.

Computation of Connection Matrix: Estimating the network con-

nectivity is commonly employed by correlation and partial correla-

tion to quantify brain activity between the ROIs. These methods

provide similarity information between the time series or frequency

spectra of nodes which can be used to construct the connectivity

matrix C. The wiring diagram of the brain regions obtained in this

manner is commonly called the connectome which is formed by the

matrix representation of all pairwise connections between ROIs.

3.

Thresholding: This is the process of filtering the connectivity ma-

trix C such that connectivity values below a certain parameter are

deleted from this matrix. As a result, the connectivity matrix C

is processed to yield a binary matrix A such that entry aij = 1 if

node i is connected to node j. A fixed threshold or a fixed threshold

node degree or a fixed edge density value may be used to filter the

connection matrix [2].

The processing of these steps that results in the connectome of the brain is

depicted in Figure 9.1. The connection matrix C provides the representation

of an edge weighted graph G = (V, E, w) where V is the set of nodes, E is

the set of edges and function f : E →R provides the weights associated with

edges which can be directed or undirected, depending on the interpretations of

interactions between the nodes. This graph may be considered as a complete

graph by associating a null value for an edge between two nodes that are not

related. The binary adjacency matrix A can be used to build an unweighted,

undirected graph G = (V, E) over which various analysis methods may be

applied.