Authors :
George Karraz
Volume/Issue :
Volume 8 - 2023, Issue 1 - January
Google Scholar :
https://bit.ly/3IIfn9N
Scribd :
https://bit.ly/3DQWXUq
DOI :
https://doi.org/10.5281/zenodo.7620853
Abstract :
A Lot of medical projects aim to combine
biology with computer science like artificial limb which
is able to simulate real limb's activities to some extent,
and that requires to comprehend the neurological map of
the brain. The best way to measure the brain's activity is
Functional Magnetic Resonance Imaging (fMRI), where
it is a functional neuroimaging procedure using MRI
technology that measures brain activity by detecting
changes associated with blood flow. In this paper we
develop an automatic system based on soft computing
methods, to analyze fMRI Images and conclude their
proper intended behavior. Our data was composed from
two parts, the major part was obtained from the famous
dataset (A test-retest fMRI dataset for motor, language
and spatial attention functions), which has a
representation of five different behaviors “finger foot
and lip movement, overt verb generation, covert verb
generation, overt word repetition and landmark tasks”,
where the second part was prepared by us using images
that free downloaded from internet network. Our
developed automatic classification system is based on
neural network framework, which is proceeding in two
stages:
1. The first stage extracts four specific features,
through applying sophisticated techniques for automatic
image processing and analysis, related to the presence of
different intensity values and their addresses over the 2
dimensions studied images. The selected features were
unique and contribute to make our system, good
represented.
2. The second stage is a classification technique,
through designing a suitable artificial intelligence system
architecture and learning algorithm. We did a lot of
experiments in order to select the best neural network
architecture and training method, the experiments
proved that the best performance was achieved in three
layers neural network: input, hidden and output layers,
with a training method based on Back propagation
algorithm, and sigmoid activation function. Developed
system achieved an accuracy of 94.4%.
Keywords :
fMRI; Neural Networks; Brain Activity Automatic Interpretation; Fuzzy C-maen clustering; Linear Regression.
A Lot of medical projects aim to combine
biology with computer science like artificial limb which
is able to simulate real limb's activities to some extent,
and that requires to comprehend the neurological map of
the brain. The best way to measure the brain's activity is
Functional Magnetic Resonance Imaging (fMRI), where
it is a functional neuroimaging procedure using MRI
technology that measures brain activity by detecting
changes associated with blood flow. In this paper we
develop an automatic system based on soft computing
methods, to analyze fMRI Images and conclude their
proper intended behavior. Our data was composed from
two parts, the major part was obtained from the famous
dataset (A test-retest fMRI dataset for motor, language
and spatial attention functions), which has a
representation of five different behaviors “finger foot
and lip movement, overt verb generation, covert verb
generation, overt word repetition and landmark tasks”,
where the second part was prepared by us using images
that free downloaded from internet network. Our
developed automatic classification system is based on
neural network framework, which is proceeding in two
stages:
1. The first stage extracts four specific features,
through applying sophisticated techniques for automatic
image processing and analysis, related to the presence of
different intensity values and their addresses over the 2
dimensions studied images. The selected features were
unique and contribute to make our system, good
represented.
2. The second stage is a classification technique,
through designing a suitable artificial intelligence system
architecture and learning algorithm. We did a lot of
experiments in order to select the best neural network
architecture and training method, the experiments
proved that the best performance was achieved in three
layers neural network: input, hidden and output layers,
with a training method based on Back propagation
algorithm, and sigmoid activation function. Developed
system achieved an accuracy of 94.4%.
Keywords :
fMRI; Neural Networks; Brain Activity Automatic Interpretation; Fuzzy C-maen clustering; Linear Regression.
