Charting developmental trajectories of dynamic brain networks during emotional face processing

Abstract

Emotional face processing is a critical component in the development of social cognition through childhood. The neural mechanisms supporting this development can be understood by tracking age-related changes in brain activity. Prior work in this area with MEG relies on static, bandlimited, or region-specific measures, which do not capture the dynamic, distributed nature of brain activity. Here we used Dynamic Network Modes (DyNeMo) to analyze MEG data from a large cohort of typically developing individuals (N=224, ages 5-40) during an emotional faces vigilance task. DyNeMo is a data-driven, generative modelling approach which captures functional networks as a set of whole-brain spatiospectral "modes" whose relative mixture (i.e., activation levels) can change dynamically in response to stimuli. We inferred six modes from the MEG data and characterized developmental trajectories in task-related activation and mode connectivity. Across modes we observed distinct developmental trajectories in both measures. With respect to mode activation, a visual mode and a frontotemporal mode (whose labels reflect their respective spatial profiles) increased nonlinearly with age; meanwhile, activation in temporal and sensorimotor modes decreased linearly with age. Meanwhile, connectivity broadly increased with age in all modes, but with different degrees of nonlinearity. These results suggest developmental dissociations between different modes (e.g., visual versus sensorimotor), as well as within individual modes (task-related mode activation versus connectivity). These results provide a comprehensive and complex picture of functional network development underlying emotional face processing. Significance StatementBrain networks supporting social cognition undergo profound changes from childhood through adolescence to adulthood. However, current understanding of how these networks develop has been limited by conventional analyses of brain imaging data, which typically provide a static picture of brain activity. Here we leveraged a cutting-edge, data-driven modeling approach (DyNeMo) to characterize age-related changes in distributed and dynamic networks supporting emotional face processing, in a large cohort of children and young adults (N=224). We inferred six functional networks whose activation levels changed rapidly in response to emotional faces. Network activation and connectivity exhibited profound and distinct non-linear changes with age, indicating that emotional face processing is supported by complex interactions among multiple dynamic networks, each with different maturational trajectories.