Author ORCID Identifier
https://orcid.org/0009-0003-2818-6759
Date of Award
5-16-2025
Degree Type
Dissertation
Degree Name
Doctor of Philosophy (PhD)
Department
Computer Science
First Advisor
Dr. Jingyu Liu
Second Advisor
Dr. Jonathan Shihao Ji
Third Advisor
Dr. Dong Hye Ye
Fourth Advisor
Dr. Jessica A Turner
Abstract
During the transition from childhood to adulthood, the human brain undergoes profound developmental changes that underpin cognitive maturation, influenced by a dynamic interplay of genetic and environmental factors. Despite the growing application of neuroimaging in brain age prediction, the specific genetic contributions to brain development remain poorly understood. This dissertation integrates genetic data with structural and resting-state functional MRI brain features to identify genetic factors influencing brain maturation and reveal linear and non-linear genetic-brain associations. Initially, we utilized multimodal brain imaging data from 1,417 participants (ages 8–22) in the PNC cohort to predict brain age. By comparing individual data modalities (e.g., gray matter density, cortical features, and functional connectivity) and their combinations, we demonstrated that traditional machine learning with feature selection performs comparably to deep learning methods. Next, we validated the replicability of brain age models in the ABCD cohort, leveraging data from 11,573 children (ages 9–10) and 2,947 children (ages 11–12). A novel refinement model focused on a narrower age range, enhanced prediction accuracy, and reliability. Using brain age estimates, we categorized 7,435 individuals (9–10 years) from the ABCD cohort into accelerated or delayed brain maturation groups. The accelerated group exhibited superior cognitive performance relative to the delayed group. To investigate environmental impacts, we conducted a multi-set canonical correlation analysis integrating brain metrics with nine environmental factors, identifying significant associations with air pollution, crime rates, and population density. Finally, to explore genetic influences, we combined structural and functional brain features with 1,030 SNPs associated with IQ, employing sparse canonical correlation analysis and deep neural networks to uncover linear and complex non-linear genetic-brain relationships. Supervised contrastive learning extracted significant latent representations, revealing decreased gray matter density in the middle temporal and inferior frontal regions, increased cerebellar and hippocampal volumes, and enhanced functional connectivity in inferior frontal and hippocampal regions in the accelerated group. The genetic analysis highlighted SNPs in genes such as CADM2, CALN1, NREP, TNRC6A, and MARK3, predominantly expressed in the frontal cortex and cerebellum. These findings provide critical insights into genetic contributions to adolescent brain maturation.
Recommended Citation
RAY, BHASKAR, "Integrating Genetic and Multimodal Brain Imaging Data for Finding Linear and Non-linear Complex Genetic-Brain Imaging Association During Brain Development." Dissertation, Georgia State University, 2025.
https://scholarworks.gsu.edu/cs_diss/234
File Upload Confirmation
1