Since the beginning of medical science, the human brain has remained an unsolved puzzle; an illusive organ that controls everything- from breathing to heartbeats, from emotion to anger, and more. With the power of advanced neuroimaging techniques, scientists have now started to solve this nearly impossible puzzle, piece by piece. Over the past decade, various in vivo techniques, including functional magnetic resonance imaging (fMRI), have been increasingly used to understand brain functions. fMRI is extensively being used to facilitate the identification of various neuropsychological disorders such as schizophrenia (SZ), bipolar disorder (BP) and autism spectrum disorder (ASD). These disorders are currently diagnosed based on patients’ self-reported experiences, and observed symptoms and behaviors over the course of the illnesses. Therefore, efficient identification of biological-based markers (biomarkers) can lead to early diagnosis of these mental disorders, and provide a trajectory for disease progression. By applying advanced machine learning techniques on fMRI data, significant differences in brain function among patients with mental disorders and healthy controls can be identified. Moreover, by jointly estimating information from multiple modalities, such as, functional brain data and genetic factors, we can now investigate the relationship between brain function and genes.
Functional connectivity (FC) has become a very common measure to characterize brain functions, where FC is defined as the temporal covariance of neural signals between multiple spatially distinct brain regions. Recently, researchers are studying the FC among functionally specialized brain networks which can be defined as a higher level of FC, and is termed as functional network connectivity (FNC, defined as the correlation value that summarizes the overall connection between brain ‘networks’ over time). Most functional connectivity studies have made the limiting assumption that connectivity is stationary over multiple minutes, and ignore to identify the time-varying and reoccurring patterns of FNC among brain regions (known as time-varying FNC).
In this dissertation, we demonstrate the use of time-varying FNC features as potential biomarkers to differentiate between patients with mental disorders and healthy subjects. The developmental characteristics of time-varying FNC in children with typically developing brain and ASD have been extensively studies in a cross-sectional framework, and age-, sex- and disease-related FNC profiles have been proposed. Also, time-varying FNC is characterized in healthy adults and patients with severe mental disorders (SZ and BP). Moreover, an efficient classification algorithm is designed to identify patients and controls at individual level. Finally, a new framework is proposed to jointly utilize information from brain’s functional network connectivity and genetic features to find the associations between them. The frameworks that we presented here can help us understand the important role played by time-varying FNC to identify potential biomarkers for the diagnosis of severe mental disorders.
fMRI, Schizophrenia, Mental illness, Independent component analysis, Functional network connectivity, machine learning
Level of Degree
Electrical and Computer Engineering
First Committee Member (Chair)
Dr. Vince D. Calhoun
Second Committee Member
Dr. Marios S. Pattichis
Third Committee Member
Dr. Meeko Oishi
Fourth Committee Member
Dr. Erik B. Erhardt
Rashid, Barnaly. "Approaches For Capturing Time-Varying Functional Network Connectivity With Application to Normative Development and Mental Illness." (2016). http://digitalrepository.unm.edu/ece_etds/314
Available for download on Monday, December 17, 2018