Longitudinal Study of Adolescent Brain Connectivity Development Using Sign-Aware Graph Theory Metrics

Abstract

Adolescence is marked by significant changes in brain network organization that underlie cognitive and behavioral development. The sensorimotor-association (SA) axis has been proposed as a hierarchical framework for understanding functional connectivity development, but most studies rely on cross-sectional data and treat positive and negative connections equivalently. We analyzed longitudinal resting-state fMRI data from 125 adolescents who passed quality control of a total of 151 (ages 12–18, 364 total scan sessions across three time points) using both functional connectivity strength and graph-theoretical metrics, comparing results from absolute-value networks (collapsing connection signs) versus sign-aware approaches. Functional connectivity strength showed age-related changes following the SA axis selectively for positive connections (r = −0.614, p < 0.001), with stronger effects in sensorimotor regions, while negative connections showed no SA alignment (r = 0.031, p = 0.803). Critically, graph-theoretical measures revealed opposing developmental gradients depending on network construction: clustering coefficient and local efficiency showed association-dominant patterns in absolute-value networks (r = 0.317, p < 0.001; r = 0.427, p = 0.001) but sensorimotor-dominant patterns in positive-only networks (r = −0.225, p < 0.001; r = −0.277, p < 0.001). Participation coefficient, an integration-based measure, showed no significant SA association in either construction. These findings demonstrate that developmental inferences critically depend on how negative connections and network topology are treated, challenging the notion of a single organizational gradient and highlighting the necessity of sign-aware graph-theoretical approaches for understanding adolescent brain maturation.