Atypical Development of Functional Brain Networks in Neonates with Congenital Heart Disease

408 

Abstract Submission 

Jung-Hoon Kim¹, Josepheen De Asis-Cruz¹, Nickie Andescavage¹, Adre du Plessis¹, Catherine Limperopoulos¹

¹Children's National Hospital, Washington, DC

Jung-Hoon Kim, PhD  
Children's National Hospital
Washington, DC

Josepheen De Asis-Cruz  
Children's National Hospital
Washington, DC

Nickie Andescavage, MD  
Children's National Hospital
Washington, DC

Adre du Plessis, MD  
Children's National Hospital
Washington, DC

Catherine Limperopoulos, PhD  
Children's National Hospital
Washington, DC

A major risk factor for newborns with congenital heart disease (CHD) is delayed brain development, enduring long-term neuromotor and neurocognitive deficits [1,2]. Previous studies have suggested neurobehavioral dysfunction in CHD is already evident in the neonatal period [3,4]. Our previous work has suggested that newborns with CHD showed aberrant brain circuitry [5]. Up to now, however, it remains unknown which large-scale brain networks corresponding to different neurocognitive functions, e.g., visual, somatosensory, auditory, or attention, are vulnerable in CHD. In this work, by leveraging two population-scale datasets acquired from independent sites, we investigated atypical development of brain networks in high-risk CHD prior to open heart surgery compared to healthy controls.

We analyzed two resting-state functional MRI (rsfMRI) datasets; one public dataset (dHCP dataset) from developing Human Connectome Project consisting of 167 good quality rsfMRI scans of healthy neonates (postmenstrual age, PMA≥41wks; PMA=42.46±1.10wks) [6] and another dataset, acquired at our institute, consisting of 137 healthy neonates (PMA=41.71±1.78wks) and 74 neonates with CHD (PMA=39.96±1.83wks). The dHCP dataset was utilized to define the normative set of functional brain networks (FBNs) with melodic ICA [7]. The number of FBNs was set to 15 heuristically while consistency over varying # of FBNs (=10, 20, and 30) was observed. Our dataset was used for investigating the difference in FBNs between healthy control and CHD. The reconstruction degree was measured by averaging squared correlation between original- and reconstructed cortical patterns using entire or individual FBN(s) over whole timepoints. The occurrence frequency of each FBN per scan was calculated by assigning volume to one of 15 FBNs given minimal cosign distance and dividing by the total volumes. Finally, the group-specific FBN maps were derived by using dual regression.

Fifteen FBNs defined from the healthy newborn population represented various brain networks spanning sensory-related regions (Figure 1A, sensorimotor: FBN1, 2, 3, and 5; visual: 9 and 13, auditory: 7 and 10). Some FBNs covered trans-modal regions, e.g., 6, 7, 8, 11, 12, 14, and 15. We observed the reconstruction degree increased over aging in healthy controls (Figure 1B; r=0.47, p<10^-7), but this was not observed in the CHD (r=0.18, p=0.12). CHD showed worse reconstruction level than control when considering age as a covariate (top row, Figure 1C). Interestingly, the association of reconstruction level to age and CHD was distinct across different FBNs (Figure 1D). For example, FBN1 covering sensorimotor regions varied significantly by CHD but not by age. Conversely, the reconstruction degree of FBN7 spanning auditory-associated regions was associated with age but not with CHD.
Next, we further investigated group-wise difference in FBN1 (CHD vs. age-matched control; n=45, PMA<41wks). We found that the occurrence of FBN1 was greater in controls than CHD group (Figure 2B, control; 11.55±3.36% vs. CHD; 8.03±5.21%, p<10^-4) while both groups shared similar patterns compared to the normative FBN pattern (Figure 2A; pattern similarity in CHD, r=0.83 vs. in control, r=0.84).

We report that newborns with CHD exhibit atypical brain activity patterns compared to healthy newborns prior to open heart surgery. Notably, unimodal brain networks, specifically related to sensorimotor system, were less represented in the CHD group. Interestingly, while both CHD and healthy controls share similar patterns of sensorimotor network, the occurrence of this network was less frequent in CHD. Our findings provide novel insights into the adverse effect of CHD on functional brain networks. In future works, we plan 1) to investigate the diversity of FBNs across subtypes of CHD and 2) to relate the variations of FBNs in CHD with neurodevelopmental outcomes.

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

Normal Brain Development: Fetus to Adolescence ²

fMRI Connectivity and Network Modeling

Congenital

FUNCTIONAL MRI

Pediatric Disorders