F. Gonzalezlima et Ar. Mcintosh, ANALYSIS OF NEURAL-NETWORK INTERACTIONS RELATED TO ASSOCIATIVE LEARNING USING STRUCTURAL EQUATION MODELING, Mathematics and computers in simulation, 40(1-2), 1995, pp. 115-140
Brain imaging techniques have the potential of providing information a
bout functional interactions within entire neural networks. Large quan
tities of data can be obtained from mapping studies, but computational
techniques are needed to make sense of the complex network interactio
ns that take place in the brain. Structural equation modeling may prov
ide such a technique by combining the anatomical connectivity with the
covariation in the activity between brain regions. Functional strengt
hs of anatomical connections between the structures that form a neural
network can be quantified by assigning numerical values to the links.
Changes in these values are used as indices of how information is pro
cessed and modified within the brain in a given situation. We used bra
in metabolic data from auditory learning experiments to explain how st
ructural models of the auditory system reveal the patterns of network
interactions related to opposite learned associative properties of the
same sound. This analysis supports the hypothesis that associative le
arning is an emergent network property, distributed among interacting
brain regions. Understanding such a property requires a network analys
is of the patterns of interactions between brain regions, rather than
the traditional analysis of regions one at a time.