Predictive functional network connectivity signature of infancy age and eye-region fixation behavior

Abstract

Functional connectivity (FC) has been established as a fundamental property of brain organization, intricately linked to both neurodevelopmental processes and the emergence of social behavior. Prior research has demonstrated that individual variability in brain age is associated with cognitive performance, primarily through the architecture of brain networks, underscoring the predictive ability of FC to characterize brain maturation. However, most existing work has focused on later developmental stages across the lifespan, with comparatively limited attention to early infancy—a period marked by rapid neurobiological growth and the emergence of foundational social skills. In this study, we applied an independent component analysis (ICA) framework to resting-state fMRI data from a longitudinal infant cohort, spanning from birth to 8 months of age, to characterize functional network connectivity (FNC). Using a linear mixed-effects model, we identified widespread but spatially heterogeneous associations between FNC and chronological age (maximum r = 0.6420, FDR-corrected p < 0.05), indicating region-specific developmental trajectories. To further model multivariate patterns of neurodevelopment, we employed partial least squares regression (PLSR) to predict brain age from whole-brain FNC. The model demonstrated high predictive accuracy (r = 0.8704, permutation test p < 1.0 × 10 ⁻3 ), confirming the feasibility of connectome-based brain age estimation in early infancy. Predictive performance remained robust in a subset of infants identified as low likelihood for autism spectrum disorder (r = 0.8713, permutation test p < 1.0 × 10 ⁻3 ). A model incorporating gestational age (i.e., corrected age) showed similarly significant performance (r = 0.8902, permutation test p < 1.0 × 10 ⁻3 ) and yielded a modest but significant improvement in accuracy (two-sample t-test p = 2.57 × 10 ⁻153 ), highlighting the importance of perinatal timing in neurodevelopmental modeling. In addition to brain age, FNC also predicted individual differences in early social attention, specifically fixation to the eyes (r = 0.4684) and mouth (r = 0.4687), with both associations reaching statistical significance (p < 1.0 × 10 ⁻3 ). Notably, predictive maps for eye and mouth fixation exhibited a robust inverse correlation (r = -0.5016, p = 2.08 × 10 ⁻145 ), suggesting that attentional allocation toward these socially salient features may be driven by distinct and potentially antagonistic patterns of brain network organization, implicating early adaptive social signaling mechanisms. Collectively, these findings suggest that resting-state FC, combined with multivariate modeling approaches, may capture meaningful associations with individual differences in brain maturation and early social behavior during infancy. The present work highlights the potential utility of functional connectome-based frameworks for studying early neurodevelopmental variability and characterizing developmental processes during critical periods of brain maturation.