THE 3
rd
INTERNATIONAL SCIENTIFIC CONFERENCES OF STUDENTS AND YOUNG RESEARCHERS
dedicated to the 99
th
anniversary of the National Leader of Azerbaijan Heydar Aliyev
94
APPLICATION OF A NEURAL NETWORK IN
ELECTROGRASTROGRAPHY
Gunel Agayeva
Azerbaijan Technical University
Baku, Azerbaijan
gunel_asoa@gmail.com
Supervisor: Associate Professor Namik Abdullaev
Keywords:
neural
network, electrogastrography,
classification of diseases, gastric
ulcer. modular structure.
Informative signs obtained to study the functional status of the organs of
the gastrointestinal tract and to classify diseases can be used as input
parameters of the artificial neural network. [1]
Topological analysis of artificial neural networks, analysis of structures,
analysis of activation functions, teaching methods and algorithms, as well as
the study of methods for normalization of input parameters justify the choice
of a multilayer perceptron scheme. Consistent study of the characteristics of
the input images through a multilayer perceptron scheme and more efficient
classification are noted as positive features of this structure.
The analysis of
various algorithms confirms the advantage of the algorithm "Back
propagation" during the operation of the learning phase of the artificial neural
network. The neural network learning technology with the teacher is more
suitable for the classification of input images. The structure of the multilayer
perceptron for the study of normal and ulcerative diseases of the
gastrointestinal tract One of the most important problems in such structures
is to determine the number of artificial neurons in the intermediate layers,
because the metrological parameters of the neural network (accuracy and
generalized capacity of the system) depend on this value. In
electrophysiological studies, a structure of a modular neural network based
on a multilayered perceptron has been proposed to classify the functional
status of organs.
The structure of the proposed modular-type artificial neural network to
study the functional status of the gastrointestinal tract
is presented in Figure
1. Fourier complex coefficients (Gi) are used as input parameters. It should
be noted that the effective (Pi) value of the electrogastroenterographic signal
can also be used as an additional input parameter to minimize the error of
the diagnostic operation [2 ]. The main advantage of the modular structure is
that each module is focused on the diagnosis of a disease, which reduces
the likelihood of misdiagnosis and increases the accuracy of the whole
system. On the other hand, the introduction of a modular structure increases
the functional capacity of the artificial neural network as a result of the