All ETDs from UAB

Advisory Committee Chair

Rajesh Kana

Advisory Committee Members

Sarah O'Kelley

Kristina Visscher

Document Type

Thesis

Date of Award

2017

Degree Name by School

Master of Arts (MA) College of Arts and Sciences

Abstract

Standard fMRI studies of healthy as well as clinical populations rely heavily on group-level averages to draw inferences about brain and behavior. This presumes neural and behavioral homogeneity within diagnostic groups resulting in group-level models which may not capture individual variability. This problem is especially relevant for studies of neurodevelopmental disorders such as autism spectrum disorder (ASD) and attention deficit hyperactivity disorder (ADHD) which clinically present with widespread individual differences. The prevalence of comorbidity between these disorders is roughly 28% highlighting the possibility that there may be shared behavioral and neural markers which cut across the diagnostic boundaries delineating the disorders. In order to better characterize these similarities and differences, the current study used a number of data-driven analyses which were blind to diagnostic classification to derive clusters of participants based on functional connectivity among regions within known brain networks. In the first approach, a unified structural equation modeling technique (Group Iterative Multiple Model Estimation) was used to derive subgroups based on connections among regions within three core brain networks: Default Mode, Salience, and Executive Control. For the second analysis, an independent component analysis was used to define components of functionally correlated brain networks. Voxel intensity values were extracted from the component maps of each participant and used in a community detection analysis to identify the community structure of the participants based on deviations from the group-level components. For the first analysis, 2-3 heterogeneous subgroups were identified for each network, but limitations inherent to the data prevented more robust, generalizable results. Only one cluster was derived from the community detection algorithm, but a number of outlier participants were identified. These results indicate that heterogeneity is a core concern in fMRI analyses and, despite limitations with the data, the analyses presented here provide a useful framework for researchers attempting to conduct neuroimaging analyses that account for individual variability.

Share

COinS