INDIVIDUAL-SPECIFIC NETWORK-LEVEL AND AREAL-LEVEL PARCELLATIONS OF THE HUMAN CEREBRAL CORTEX

Abstract

Resting-state functional magnetic resonance imaging (rs-fMRI) offers the opportunity to derive parcellations of the human cerebral cortex in vivo. Until recently, most parcellations have relied on data averaged across many individuals. However, such group-level parcellations could obscure individual-specific features that are biologically meaningful. In this thesis, I propose a multi-session hierarchical Bayesian model (MS-HBM) for estimating network-level and areal-level individual-specific parcellations. The multiple layers of the MS-HBM explicitly differentiate intra-subject (within-subject) from inter-subject (between-subject) variability. By explicitly modelling intra-subject variability, the MS-HBM avoids the common pitfall of potentially confusing intra-subject variability for inter-subject differences. Compared with other approaches, MS-HBM network-level and areal-level parcellations generalized better to new rs-fMRI and task-fMRI data from the same subjects. Importantly, I show that behavioral phenotypes across domains of cognition, personality and emotion could be predicted by topography, size and connectivity strength from individual-specific parcellations.