Diagnosing ADHD by MR images using meta-cognitive radial basis function network
The purpose of this experiment is to explore two different feature selection methods, the T-test and Spectral Feature Selection, on the training data, so that the features that are more crucial and contribute most to detecting whether a child has ADHD can be extracted and used to train the Meta-cogn...
Saved in:
Main Author: | |
---|---|
Other Authors: | |
Format: | Final Year Project |
Language: | English |
Published: |
2014
|
Subjects: | |
Online Access: | http://hdl.handle.net/10356/59007 |
Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
Institution: | Nanyang Technological University |
Language: | English |
id |
sg-ntu-dr.10356-59007 |
---|---|
record_format |
dspace |
spelling |
sg-ntu-dr.10356-590072023-03-03T20:27:23Z Diagnosing ADHD by MR images using meta-cognitive radial basis function network Praveena Satkunarajah School of Computer Engineering Centre for Computational Intelligence Ast/P Suresh Sundaram DRNTU::Engineering::Computer science and engineering::Computing methodologies::Artificial intelligence The purpose of this experiment is to explore two different feature selection methods, the T-test and Spectral Feature Selection, on the training data, so that the features that are more crucial and contribute most to detecting whether a child has ADHD can be extracted and used to train the Meta-cognitive Radial Basis Function Network (McRBFN). Feature selection helps to reduce the dimensionality of the data and sheds features that are irrelevant to the learning process. The McRBFN is a neural network which aims to discover a function which maps training sample data to their correct classes. By doing this, it may be possible to diagnose whether a child has ADHD from the child’s Magnetic Resonance (MR) Images of his brain. The training data was obtained from ADHD-200 consortium data set and then processed by Voxel Based Morphometry to extract regions of interest, which in this experiment was the amygdala region of the brain. Both feature selection methods were used to rank the features. The first ten features from each of the rankings were extracted from the 1050 features in the data and run through the McRBFN. The number of features was incremented by 10 until the results of the change in results for the overall and average training and testing frequencies became smaller, after which the number of features were incremented by 5 until the results stagnated. Bachelor of Engineering (Computer Science) 2014-04-21T02:32:49Z 2014-04-21T02:32:49Z 2014 2014 Final Year Project (FYP) http://hdl.handle.net/10356/59007 en Nanyang Technological University 30 p. application/pdf |
institution |
Nanyang Technological University |
building |
NTU Library |
continent |
Asia |
country |
Singapore Singapore |
content_provider |
NTU Library |
collection |
DR-NTU |
language |
English |
topic |
DRNTU::Engineering::Computer science and engineering::Computing methodologies::Artificial intelligence |
spellingShingle |
DRNTU::Engineering::Computer science and engineering::Computing methodologies::Artificial intelligence Praveena Satkunarajah Diagnosing ADHD by MR images using meta-cognitive radial basis function network |
description |
The purpose of this experiment is to explore two different feature selection methods, the T-test and Spectral Feature Selection, on the training data, so that the features that are more crucial and contribute most to detecting whether a child has ADHD can be extracted and used to train the Meta-cognitive Radial Basis Function Network (McRBFN). Feature selection helps to reduce the dimensionality of the data and sheds features that are irrelevant to the learning process.
The McRBFN is a neural network which aims to discover a function which maps training sample data to their correct classes. By doing this, it may be possible to diagnose whether a child has ADHD from the child’s Magnetic Resonance (MR) Images of his brain. The training data was obtained from ADHD-200 consortium data set and then processed by Voxel Based Morphometry to extract regions of interest, which in this experiment was the amygdala region of the brain.
Both feature selection methods were used to rank the features. The first ten features from each of the rankings were extracted from the 1050 features in the data and run through the McRBFN. The number of features was incremented by 10 until the results of the change in results for the overall and average training and testing frequencies became smaller, after which the number of features were incremented by 5 until the results stagnated. |
author2 |
School of Computer Engineering |
author_facet |
School of Computer Engineering Praveena Satkunarajah |
format |
Final Year Project |
author |
Praveena Satkunarajah |
author_sort |
Praveena Satkunarajah |
title |
Diagnosing ADHD by MR images using meta-cognitive radial basis function network |
title_short |
Diagnosing ADHD by MR images using meta-cognitive radial basis function network |
title_full |
Diagnosing ADHD by MR images using meta-cognitive radial basis function network |
title_fullStr |
Diagnosing ADHD by MR images using meta-cognitive radial basis function network |
title_full_unstemmed |
Diagnosing ADHD by MR images using meta-cognitive radial basis function network |
title_sort |
diagnosing adhd by mr images using meta-cognitive radial basis function network |
publishDate |
2014 |
url |
http://hdl.handle.net/10356/59007 |
_version_ |
1759854533888966656 |