A0460
Title: Disentangled adversarial flow with ensemble learning for multi-source brain connectome analysis
Authors: Yixin Chen - Virginia Tech (United States)
Zhengwu Zhang - UNC Chapel Hill (United States)
Xin Xing - Virginia Tech University (United States)
Meimei Liu - Virginia Tech (United States) [presenting]
Abstract: Understanding the brain's structural connectome and its role in cognitive functions has been advanced by diffusion magnetic resonance imaging. However, the heterogeneity across multiple neuroimaging studies, combined with limited labeled samples in specialized cohorts, poses significant challenges in developing accurate predictive models for cognitive abilities. A novel disentangled adversarial flow (DAF) model is introduced, leveraging large-scale datasets to enhance predictions in smaller neuroimaging studies. DAF generates domain-invariant brain connectome representations using a bidirectional architecture and a kernel-based measure to minimize domain label dependence. An ensemble DAF regression framework integrates multiple data sources, reducing information loss in multi-domain data. Validated on the adolescent brain cognitive development (ABCD) study, the human connectome project (HCP), and the Alzheimer's disease neuroimaging initiative (ADNI), DAF shows reduced discrepancies across domains and superior predictive accuracy with limited target domain data.
