A Software for Multivoxel Pattern Analysis of Functional Magnetic Resonance Imaging Data

Kushal Bohra

Document Type

Thesis

Degree Name

Master of Science (MS)

Department

Computer Science and Info Sys

Date of Award

Summer 2014

Abstract

Functional magnetic resonance imaging (fMRI) is a medical imaging modality which enables taking 3D snapshots of the whole brain usually every few seconds with approximately 3-30 millimeter-cube resolution. FMRI signal is a blood-oxygenation-level-dependent (BOLD) signal, which is an indirect measure of the brain's neural activity. This technique relies on the fact that cerebral blood flow and neuronal activation are coupled. Exposing the subject to different kind of stimuli or tasks systematically during an fMRI scan causes different parts of the brain to activate. FMRI signals can be analyzed to find out the characteristics of these activations and any systematic differences between the activations under different stimuli and tasks. Multivariate pattern recognition and classification algorithms, also known as multivoxel pattern analysis (MVPA) in the neuroimaging community, have been recently applied to fMRI data and it has been gaining popularity. MVPA of fMRI data have allowed classification of different brain states under different stimuli or task conditions. For example, it is possible to distinguish, based on the fMRI data alone, whether a subject was looking at faces vs. tools, tools vs. houses etc. with a good specificity and sensitivity. However, the MVPA functions or tools that have been developed specifically for fMRI are not user-friendly. Data analysis require significant coding/programming by researchers, who are mostly psychologists, neurologists or neuroscientists. In this work, we developed a user-friendly, MATLAB-based GUI toolbox that allows fast, efficient and intuitive analysis of fMRI data. Many pattern recognition/classification functions, which are written for general 2-dimensional data, already exist in MATLAB. These functions were expanded to include the latest pattern recognition/classification analysis algorithms such as support-vector machines with multi-class support. These and already existing functions were tailored to analyze the fMRI data, which are 4-dimensional (space-time) and stored in a different data format (NIfTI-1/.nii) other than MATLAB's data format (.mat). We tested our toolbox by doing a complete analysis of the six-subject fMRI dataset by Haxby et al., 2001 "Faces and Subjects in Ventral Temporal Cortex" and compared classification results from different classification algorithms.

Advisor

Unal Sakoglu

Subject Categories

Computer Sciences | Physical Sciences and Mathematics

Link to Full Text

Share

COinS