Document Type

Thesis

Date of Award

Fall 1-27-2008

Degree Name

Master of Science in Biomedical Engineering - (M.S.)

Department

Biomedical Engineering

First Advisor

Tara L. Alvarez

Second Advisor

Bharat Biswal

Third Advisor

Richard A. Foulds

Abstract

The development of fMRI (functional Magnetic Resonance Imaging) has led many researchers to localize brain functions using different stimuli. The use of pattern recognition techniques have made it possible to predict the stimuli being presented from the corresponding brain images and activation patterns. The primary objective of the present study was to use pattern recognition methods to develop a model using available fMRJ images and then to use the model to identify the stimulus presented from a large number of unknown images. Two different experimental conditions were used involving both binary and multi-class classification. Bilateral finger tapping data which had two distinct states "Active" and "Rest" were used for binary classification. Binary classification was done using Learning Vector Quantization (LVQ) and Least Square Support Vector Machine (LS-SVM). Gas mixture data, which were obtained from rats while ventilated with different gas mixtures for rest and breath hold task, gave various physiological conditions. These multi-class data were also classified using LS-SVM technique. Feature selection was performed on every data to select out patterns made up of significant voxels using statistical techniques like correlation, paired t-test and ANOVA. The accuracies for binary classification were between 90% and 100% while the average accuracy for multi-categorical data was 70%.

Share

COinS
 
 

To view the content in your browser, please download Adobe Reader or, alternately,
you may Download the file to your hard drive.

NOTE: The latest versions of Adobe Reader do not support viewing PDF files within Firefox on Mac OS and if you are using a modern (Intel) Mac, there is no official plugin for viewing PDF files within the browser window.