Early Development of Brain Functional Networks

377 

Abstract Submission 

Kimhan Thung¹, WENJIAO LYU¹, Li Wang¹, Weili Lin¹, Sahar Ahmad¹, Pew-Thian Yap¹

¹The University of North Carolina at Chapel Hill, Chapel Hill, NC

Kimhan Thung  
The University of North Carolina at Chapel Hill
Chapel Hill, NC

WENJIAO LYU  
The University of North Carolina at Chapel Hill
Chapel Hill, NC

Li Wang  
The University of North Carolina at Chapel Hill
Chapel Hill, NC

Weili Lin  
The University of North Carolina at Chapel Hill
Chapel Hill, NC

Sahar Ahmad  
The University of North Carolina at Chapel Hill
Chapel Hill, NC

Pew-Thian Yap  
The University of North Carolina at Chapel Hill
Chapel Hill, NC

We investigated the development of baby brain functional networks from birth to 5 years old. Using over 1,200 resting state functional MRI (rs-fMRI) scans from the Baby Connectome Project (BCP), we quantified functional development during early childhood, covering sensorimotor, visual, auditory, default mode, and cerebellar networks.

We used rs-fMRI data of 280 subjects scanned in the BCP [1]. Preprocessing of rs-fMRI data [2] includes head motion correction, EPI distortion correction, fMRI to structural MRI registration guided by tissue segmentation maps, one-time resampling of fMRI data in subject native space, high-pass filtering, and ICA-AROMA denoising. We then independently performed group ICA (30 components) for each month (based on time windows defined in [4]) to obtain month-specific functional networks. For temporal consistency, we further performed group ICA (35 components) on these month-specific ICA components to obtain an overall template of functional networks, which were in turn used to compute the functional networks at each month via dual regression [3]. For each functional network, we fitted a Generalized Additive Mixed-effect Model (GAMM) to each voxel: Y ~ s(age) + (1|RID) + (1|site), where s(.) is a smooth function, and (1|.) represents site and subject random effects. The developmental trajectory for each network was plotted for a high-activation point. We compared the development patterns of 5 primary function networks (i.e., 2 sensorimotor, 1 auditory, and 2 visual networks) and 5 higher order association networks (i.e., cerebellar, 2 default-mode networks, and 2 executive control networks).

The development of functional networks can be observed from the spatial maps and the activation trajectory curves. Generally, all functional networks emerge since birth and stabilize after month 6. The activation increases substantially from birth to month 6 and increases gradually through month 60. The activations of association networks are in general substantially lower than the primary function networks.

Fig. 1: Month-specific functional atlases for major resting state functional networks: (a) sensorimotor networks (left to right) - motor networks (SMN-lateral, SMN-medial), auditory network, and visual networks (VIS-occipital, VIS-medial); (b) association networks (left to right) - cerebellar network, default mode networks (DMN-prefrontal, DMN-Precuneus), and executive control networks (ECN-L, ECN-R). The rows correspond (top to bottom) to different months: 0, 1, 3, 9, 12, 24, and 60.

Fig. 2. Developmental trajectories of functional networks. Shaded areas mark the standard errors.

We constructed monthly functional atlases from birth to five years of age, capturing the spatiotemporal characteristics of early brain development.

Neurodevelopmental/ Early Life (eg. ADHD, autism) ¹

Early life, Adolescence, Aging ²

Activation (eg. BOLD task-fMRI)

Aging

FUNCTIONAL MRI

PEDIATRIC