A0943
Title: Learning directed brain connectomes for neurodevelopment
Authors: Ying Guo - Emory University (United States) [presenting]
Abstract: In recent years, connectome-based research has emerged as a central focus in neuroscience, offering critical insights into brain organization and supporting predictive modeling of cognitive, behavioral, and clinical outcomes. While most existing approaches analyze undirected brain connectomes, they overlook the directionality and causal influence between brain regions. To address this limitation, the aim is to propose a novel, task-aware framework for learning directed brain connectomes from fMRI data. The method leverages advanced statistical modeling and machine learning to perform regularized inference, yielding sparse, directed connectivity graphs that capture causal interactions across the brain. Simultaneously, the method learns low-dimensional graph embeddings that are optimized to predict demographic, behavioral, and clinical outcomes. Applied to a large-scale neurodevelopmental study, the approach uncovers directed whole-brain connectivity patterns among children and adolescents and reveals new insights into subpopulation differences in the directed connectome, highlighting its potential to advance both mechanistic understanding and predictive modeling in neuroscience.
