Date of Award

Spring 2006

Document Type

Thesis

Degree Name

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

Department

Biomedical Engineering

First Advisor

Bharat Biswal

Second Advisor

Tara L. Alvarez

Third Advisor

Ronald H. Rockland

Abstract

Autism is a neurobiological disorder in which, certain regions of the brain are affected. The main features of autism are impairment in communication, social interaction, language and deficit in imitation and theory of mind. Using Functional Magnetic Resonance Imaging (fMRI), haemodynamic responses during a bilateral finger tapping task are analyzed for both autistic subjects and normal control subjects. fMRI is a noninvasive technique to image the activity of the brain related to a specific task. Generally, the active voxels in the IMRI images are detected using parametric or non-parametric statistical methods in which the fMRI response is assumed to have a model. Such methods are not applicable to detect the active voxels when the fMRI response is unknown. The data driven methods are also used for analyzing the fMRI data. The data driven methods are computationally expensive.

In this study, a method for detecting activated voxels without using prior knowledge of the input stimulus is presented. The assumption in this method is that the activation typically involves larger region comprising of several voxels and that these neighboring activated voxels are also temporally correlated. To validate the accuracy of this method, Principal component Analysis and Independent Component Analysis are also performed. A significant overlap in the sensorimotor cortex is found between the various methods suggesting that the automatic detecting method presented does provide accurate detection and localization.

Share

COinS