From Geometry to Hierarchy: Charting the Dominant Modes of Human Cortical Function in Development

1760 

Abstract Submission 

Alexander Holmes¹, James Pang¹, Alex Fornito²

¹Monash University, Melbourne, Victoria, ²Monash University, Clayton, Victoria

Alexander Holmes  
Monash University
Melbourne, Victoria

James Pang, PhD  
Monash University
Melbourne, Victoria

Alex Fornito  
Monash University
Clayton, Victoria

Low-dimensional representations of brain-wide inter-regional functional coupling (FC) offer a powerful tool for uncovering dominant organizational motifs of the cerebral cortex. Neural field theory, a well-validated, biophysically constrained framework for modelling large-scale brain activity, predicts that the dominant spatial eigenmodes of cortical dynamics should correspond to the modes of cortical geometry, under the assumption that neural dynamics are dominated by distance-dependent wave-like propagation (Robinson et al., 2016). In line with this view, recent work has shown that eigenmodes of cortical geometry offer a parsimonious explanation for how anatomy constrains diverse aspects of brain function (Pang, et al., 2023), but a precise alignment between these geometric modes and the dominant modes of dynamics is lacking. For instance, the first non-global mode of geometry follows an anterior-posterior (A-P) gradient, whereas the dominant mode of inter-regional FC estimates is organized along a sensorimotor-association (S-A) axis that reflects classically described cortical processing hierarchies (Margulies et al., 2016; Mesulam, 1998). Notably, while this discrepancy is apparent in adult brains, it is less salient in neonates, where geometric and functional modes show greater alignment (Lariviere et al., 2020), suggesting that maturation of complex patterns of connectivity that perturb simple distance-dependent processes may lead to the gradual emergence of an S-A axis from a functional organization that is initially dominated by geometry. Here, we tested this hypothesis using three independent fMRI datasets spanning age ranges from birth to 21 years.

We obtained fMRI data from 282 individuals in the UNC/UMN Baby Connectome Project (BCP; aged 1-60 months), 650 individuals in the Human Connectome Project – Development cohort (HCP-D; aged 6-21 years), and 347 individuals in the Nathan Kline Institute – Rockland Sample (NKI-RS; aged 6-21 years). Across 1,279 total participants, we computed low-dimensional connectivity manifolds via diffusion map embedding using FC matrices extracted from 11,524 cortical surface vertices, with group average manifolds also obtained across representative age bands. Geometric eigenmodes were calculated using the Laplace-Beltrami operator of a downsampled, left–right symmetric version of FreeSurfer's fsaverage population-averaged template (Pang et al., 2023). We then evaluated how well the age-specific FC modes correlated with either the canonical S-A axis obtained via diffusion map embedding in the Human Connectome Project S1200 group average FC matrix (Margulies et al., 2016), or a general linear model predicting the FC mode using the first three non-global geometric eigenmodes, which correspond to A-P, dorsal-ventral, and medial-lateral spatial patterns.

The dominant functional mode in infancy reflected an A-P pattern that transitioned towards a hierarchical S-A axis in later development (Fig. 1). The coefficient of determination of the geometric model (r²) peaked in early infancy, with a maximum value of 0.95 at 6-9 months in the BCP dataset, and declined throughout adolescence, suggesting that the proportion of variability in FC modes explained by geometric eigenmodes decreases with age (Fig. 2). The r² eventually reached a minimum value of 0.24 at 19-21 years old in the HCP-D cohort. Conversely, the r² between the S-A axis and each age-specific gradient map increased with age, peaking at 0.94 at 19-21 years old in the HCP-D cohort.

The functional organization of the infant cortex is strongly dominated by geometry and is patterned predominantly along an A-P gradient. As development progresses, this organization transitions towards a hierarchical S-A axis. These shifts may be driven by the maturation of long-range cortico-subcortical or cortico-cortical fibers.

Early life, Adolescence, Aging ²

Normal Brain Development: Fetus to Adolescence

fMRI Connectivity and Network Modeling ¹

Aging

Computational Neuroscience

Development

FUNCTIONAL MRI

Open Data