Search Papers | Poster Sessions | All Posters
Poster A29 in Poster Session A - Tuesday, August 6, 2024, 4:15 – 6:15 pm, Johnson Ice Rink
Individual brain parcellations for cognitive mapping obtained from a hierarchical Bayesian framework
Ana Luísa Pinho1,2 (), Jennifer Yoon1,2, Jörn Diedrichsen1,2,3; 1Department of Computer Science, Western University, London, Ontario, Canada, 2Western Institute of Neuroscience, Western University, London, Ontario, Canada, 3Department of Statistical and Actuarial Sciences, Western University, London, Ontario, Canada
In recent years, individual brain parcellations have become increasingly popular in human brain imaging as they provide better precision for functional localization than population-based atlases. Yet, often, there is only very little individual data available to define individual regions. Here, we exploit a Hierarchical Bayesian Parcellation (HBP) scheme to derive subject-specific parcellations extracted from a limited amount of individual task data and evaluate its performance using the Distance-Controlled Boundary Coefficient. We compare the HBP performance with Dual Regression and Dictionary Learning, two data-driven methods commonly used on resting-state and task-based data. In particular, we demonstrate that the Bayesian integration of individual data with a group prior—inferred from a large deep-behavioral phenotyping resource—provides substantial advantages in defining individual regions.
Keywords: Functional Brain Parcellation Bayesian Hierarchical Modeling Dictionary Learning Dual Regression