Identifying reproducible individual differences in childhood functional brain networks: An ABCD study

Anthony Galassi,Kathy Snider,Damien A Fair,Eric Feczko,Michaela Cordova,Darrick Sturgeon,Hugh Garavan,Eric Earl,M Deanna ,Sarah W Feldstein Ewing,Deanna J Greene,Timothy O Laumann,Robert Hermosillo,Anders M Dale,Scott Marek,Olivia Doyle,Ruth H Miller ,Alice M Graham ,Brenden Tervo‐Clemmens ,Bonnie J Nagel,Will Foran ,Muriah D Wheelock,Ashley N Nielsen,Adam T Eggebrecht,Óscar Miranda-Domínguez ,Beatríz Luna ,Anders Perrone,Nico U.F Dosenbach

doi:10.1016/j.dcn.2019.100706

Abstract

The 21-site Adolescent Brain Cognitive Development (ABCD) study provides an unparalleled opportunity to characterize functional brain development via resting-state functional connectivity (RSFC) and to quantify relationships between RSFC and behavior. This multi-site data set includes potentially confounding sources of variance, such as differences between data collection sites and/or scanner manufacturers, in addition to those inherent to RSFC (e.g., head motion). The ABCD project provides a framework for characterizing and reproducing RSFC and RSFC-behavior associations, while quantifying the extent to which sources of variability bias RSFC estimates. We quantified RSFC and functional network architecture in 2,188 9-10-year old children from the ABCD study, segregated into demographically-matched discovery (N = 1,166) and replication datasets (N = 1,022). We found RSFC and network architecture to be highly reproducible across children. We did not observe strong effects of site; however, scanner manufacturer effects were large, reproducible, and followed a “short-to-long” association with distance between regions. Accounting for potential confounding variables, we replicated that RSFC between several higher-order networks was related to general cognition. In sum, we provide a framework for how to characterize RSFC-behavior relationships in a rigorous and reproducible manner using the ABCD dataset and other large multi-site projects.

Highlights

Adolescence is a unique developmental period of specialization during which neural processes supporting goal-directed behavior continue to stabilize towards adult levels (Larsen and Luna, 2018)
In the Adolescent Brain Cognitive Development (ABCD) dataset, highly reproducible group-averaged restingstate functional connectivity (RSFC) is observed at these sample sizes
We found RSFC and network architecture was highly reproducible across children and conformed to many features observed in adult-level networks

Summary

Introduction

Adolescence is a unique developmental period of specialization during which neural processes supporting goal-directed behavior continue to stabilize towards adult levels (Larsen and Luna, 2018). Predominant neurodevelopmental models (Di Martino et al, 2014; Grayson and Fair, 2017; Luna et al, 2015) hypothesize that maturation of widely-distributed patterns of restingstate functional connectivity (RSFC) within and between functional networks underlie individual differences in cognitive development and that RSFC disruption is associated with the emergence of psychiatric disorders during adolescence. Work on person-specific patterns of RSFC suggests that the functional connectome uniquely reflects an individual's trait level features (Gratton et al, 2018) and may be predictive of individual differences in cognitive function and psychopathology. Due to challenges in recruiting neuroimaging samples representative of the full range of behavioral variability in the population, a direct and comprehensive mapping between individual variability in functional networks and trait level features has remained elusive. The ABCD dataset (Volkow et al, 2018), which has recruited close to 12,000 participants for brain imaging and behavioral assessments, affords a critical opportunity for a more systematic mapping of the associations between RSFC and behavior

Methods

Results

Discussion

Conclusion