Estimating Maternal Poverty and Predicting Infant Attention from Neonatal Functional Brain Network Connectivity

• In Event: Poster Session 06
  In Poster Session: PS 06 Topic - Biological Processes: Neuroscience and Genetics

Abstract

Prenatal poverty can have negative impacts on a range of outcomes across infancy, including early cognitive development and the structure of the underlying brain regions (Lipina & Posner, 2012; Hair et al, 2015; Nesbitt et al, 2013). However, there is a dearth of research on the impact of poverty on functional brain networks during the neonatal period, and no studies which have used machine learning to test this relationship. We will utilize multivariate analysis to test whether maternal poverty can be estimated from neonatal resting state functional brain network connectivity, and predictive of later infant attention, in an ongoing, longitudinal study of N=76 infants, (or more, depending when MRI scanning resumes). Questionnaire measures of maternal poverty were collected during a prenatal visit. Infant attention was assessed at 6 months of age with the puppet task (procedure and criteria adapted from Diamond, Prevor, Callender, & Druin, 1997; Cuevas & Bell, 2014). Infants had two consecutive, 6-minute resting state functional MRI (rs-fMRI) scans at 2 weeks of age on a Prisma 3T scanner. The rs-fMRI data processing is underway using AFNI, including motion correction, registration to template space, nuisance regression, and smoothing, and will be completed in October 2020. Statistical analysis will be completed by February 2021 using a combination of R Statistics, and sklearn and nibabel in python. Preliminary analyses indicate that at least 60 participants have 5 minutes or more of usable fMRI data with frame-to-frame motion less than 0.2 mm. Functional brain networks and nodes will be defined from an infant functional parcellation (Eggebrecht et al, 2017), which will serve as features in multivariate analysis using partial least squares regression and random sampling, similar to Rudolph et al, 2018. We hypothesize that we will accurately estimate maternal poverty from neonatal functional brain network connectivity, and the resultant models will be heavily weighted by brain regions underlying executive function. We will then test whether infant attention can be statistically predicted from these same models, which were defined by estimation of maternal poverty, or whether a separate pattern of neonatal functional network connectivity predicts infant attention. Since we hypothesize that maternal prenatal poverty affects infant attention through changes in the underlying functional brain networks, we further hypothesize that infant attention will be accurately predicted from the neural models used to estimate maternal poverty. Finally, we will also employ univariate mediation analysis to address this question, and hypothesize that functional connectivity strength within the brain network defined based on the weighted, predictive brain mask will mediate relationships between maternal poverty and infant attention. As an alternative to the multivariate-defined network, we will examine functional connectivity of the frontoparietal control and dorsal attention networks based on the template parcellation described above, as mediators of the relationship between maternal poverty and infant attention. Results of this research will elucidate neural mechanisms underlying the negative impact of poverty on infant cognitive function.

Authors

• Tammi R.A. Kral, University of Wisconsin - Madison
  Presenting Author

• Cathi Propper, University of North Carolina at Chapel Hill
  Non-Presenting Author

• Rebecca Stephens, University of North Carolina at Chapel Hill
  Non-Presenting Author

• Rasmus M Birn, University of Wisconsin - Madison
  Non-Presenting Author

• Sarah J Short, University of Wisconsin - Madison
  Non-Presenting Author