| dc.creator | JUAN MANUEL RAMIREZ CORTES | |
| dc.date | 2009-08 | |
| dc.date.accessioned | 2023-07-25T16:22:33Z | |
| dc.date.available | 2023-07-25T16:22:33Z | |
| dc.identifier | http://inaoe.repositorioinstitucional.mx/jspui/handle/1009/883 | |
| dc.identifier.uri | https://repositorioslatinoamericanos.uchile.cl/handle/2250/7806083 | |
| dc.description | This report is the result of the research developed during the summer research
stay of the first author at the Computer Vision and Image Analysis
Laboratory, Electrical and Computing Engineering Department, Texas Tech
University, USA. The first author worked with the graduate student Jingqi Ao
who obtained the results described in this report as part of his graduate studies. | |
| dc.description | Functional magnetic resonance imaging (fMRI), is a technique for investigating brain’s
activity in a mental process, in response to some specific applied stimulus. In this report, a
two-class fMRI classification project is described. fMRI (Functional Magnetic Resonance
Imaging) image is inherently high-dimensional. More than ten millions voxels in the raw
fMRI data make the fMRI-related pattern recognition difficult to perform directly. Different
methods have been proposed to solve this problem, such as raw-data related t-value1
method, Principal Component Analysis (PCA)2 and Independent Component Analysis
(ICA)3. In this project, PCA (Principal Component Analysis) is used as feature extraction
method to decrease the dimension of fMRI data. Two feature selection methods are used:
Max-group and Forward-Search. Three most commonly used classifiers in two-class
classification are compared: Support Vector Machine (SVM), Linear Discriminant Analysis
(LDA) and k-Nearest Neighbor (KNN). Results obtained from experiments using
preprocessed (high-pass filtering and motion-correction) and non-preprocessed data from 8
different subjects, are presented. Finally, two more feature extraction methods (GLM-based
t-test and Two-sample t-test) are proposed to improve the accuracy rate of classification.
Feature selection methods are kept unchanged, still containing MaxGroup and
ForwardSearch. The features from different feature extraction methods and different feature
selection methods are compared in the form of taking one subject’s result as an example.
Accuracy Rate of 8 different subjects, obtained with 4 combinations of feature extraction
methods and feature selection methods, are finally presented. | |
| dc.format | application/pdf | |
| dc.language | eng | |
| dc.publisher | Instituto Nacional de Astrofísica, Óptica y Electrónica | |
| dc.rights | info:eu-repo/semantics/openAccess | |
| dc.rights | http://creativecommons.org/licenses/by-nc-nd/4.0 | |
| dc.subject | info:eu-repo/classification/cti/1 | |
| dc.subject | info:eu-repo/classification/cti/22 | |
| dc.subject | info:eu-repo/classification/cti/2203 | |
| dc.subject | info:eu-repo/classification/cti/2203 | |
| dc.title | Feature extraction and classification of a two-class fMRI experiment using principal component analysis, LDA and SVM | |
| dc.type | info:eu-repo/semantics/report | |
| dc.type | info:eu-repo/semantics/acceptedVersion | |
| dc.audience | students | |
| dc.audience | researchers | |
| dc.audience | generalPublic | |
