B1241
Title: Supervised brain functional node and network construction related to behavior under voxel-level cognitive state fMRI
Authors: Yize Zhao - Yale University (United States)
Wanwan Xu - Yale University (United States) [presenting]
Tianxi Li - University of Michigan (United States)
Selena Wang - Yale University (United States)
Abstract: A plausible brain atlas is a prerequisite for constructing meaningful connectomics, identifying neural substrates of cognition and behavior and ultimately enhancing the understanding of the brain. On the other hand, brain parcellation is a unique application for graph theory-based methods where each region is a node in the network. The clinical insights and interpretations could lead to the development of new tools and methods. A supervised brain parcellation scheme is presented, borrowing inspiration from the spectral clustering literature. Starting from resting state or task state voxel-level fMRI data, a set of nodes for network analysis is identified. Compared to the existing atlas, the proposed method not only group the voxels with strong connections but also penalize the clustering among voxels with edges highly correlated to the cognitive outcome. The proposed method is evaluated on two datasets using connectome-based predictive modelling, where both demonstrated improved empirical results.
