Date of Award
2025
Document Type
Dissertation
Degree Name
Doctor of Philosophy (PhD)
Department
Complex Systems and Data Science
First Advisor
Nicholas Allgaier
Abstract
Task-based functional connectivity (FC) studies have gained significant attention in the field of neuroscience for their potential to unravel the neural basis of individual differences in cognitive and behavioral traits. However, the presence of high inter-individual variability in task-based FC poses challenges in establishing interpretable and generalizable findings. Although the utilization of machine learning (ML) in neuroimaging studies has shown promise in predicting cognitive abilities, personality traits, and clinical outcomes, few studies have focused on employing these approaches to identify task-based FC patterns and discriminate individuals with different cognitive profiles or clinical conditions, especially from a large-scale neuroimaging dataset. This thesis addresses this gap in knowledge, by leveraging ML algorithms and characterizing novel biomarkers in order to gain insights into the underlying mechanisms of inter-individual variability.
In Chapter 2, we performed multi-modal analysis with six different ML algorithms to estimate individual differences in executive function. Such an approach provided a generalizable and well-performed predictive model from brain activation during Stop Signal Task and highlighted a potential way to interpret its features. In Chapter 3, we applied graph theory measures to uncover the relationships between task-based FC and differences in age. It revealed that network properties from task-based FC can capture inter-individual and neurodevelopmental differences during response inhibition tasks. Last, in Chapter 4 we used task-based FC during working memory to localize brain networks associated with attention-deficit/hyperactivity disorder (ADHD) symptomatology. We demonstrated a successful ML approach for estimating ADHD symptom levels with task-based FC and showed the FC measures to perform better than regional brain activation. Results from the ADHD predictive model also provided more information about the neural correlates of inter-individual variability. In conclusion, task-based FC with ML approach offers a promising avenue for studying individual differences in cognitive and psychiatric traits. Understanding the predictive model of task-based FC has the potential to advance personalized medicine, improve cognitive training interventions, and enhance our understanding of the diverse cognitive and behavioral profiles observed across individuals.
Language
en
Number of Pages
193 p.
Recommended Citation
Yuan, Dekang, "Task-Based Functional Connectivity with Machine Learning Approaches to Detect Individual Differences in Brain Phenotype" (2025). Graduate College Dissertations and Theses. 2124.
https://scholarworks.uvm.edu/graddis/2124