Meta-Analytic Connectivity Modelling of Neural Networks for Reading and Math in Children

1034 

Abstract Submission 

Chiao-Yi Wu¹, Xiaowen Lin², SH Annabel Chen^2,3,4

¹National Institute of Education, Nanyang Technological University, Singapore, Singapore, ²Psychology, School of Social Sciences, Nanyang Technological University, Singapore, Singapore, ³Centre for Research and Development in Learning, Nanyang Technological University, Singapore, Singapore, ⁴Lee Kong Chian School of Medicine, Nanyang Technological University, Singapore, Singapore

Chiao-Yi Wu  
National Institute of Education, Nanyang Technological University
Singapore, Singapore

Xiaowen Lin  
Psychology, School of Social Sciences, Nanyang Technological University
Singapore, Singapore

SH Annabel Chen, PhD  
Psychology, School of Social Sciences, Nanyang Technological University|Centre for Research and Development in Learning, Nanyang Technological University|Lee Kong Chian School of Medicine, Nanyang Technological University
Singapore, Singapore|Singapore, Singapore|Singapore, Singapore

Literacy and numeracy are fundamental skills for the attainment of academic competence in school. Previous studies have suggested that reading and math involve both domain-specific and domain-general neurocognitive mechanisms (Ashkenazi et al., 2013). Executive function (EF) has been shown to be a domain-general ability that supports reading and math processing. While EF ability has been associated with literacy and numeracy (Willcutt et al., 2013), how it supports reading and math processing at the neurobiological level remains to be elucidated. In the current study, we performed coordinate-based activation likelihood estimation (ALE) meta-analysis and meta-analytic connectivity modelling (MACM) to examine functional connectivity between the domain-general and domain-specific networks between reading and math.

We followed the PRISMA framework (Page et al., 2021) to search for fMRI studies on reading and math in 5 databases. All selected studies (reading: 39 contrasts, 364 foci; math: 38 contrasts, 371 foci) involved typically developing children aged ≤ 13-year-old and reported MNI or Talairach coordinates from whole-brain analyses (Wu et al., 2021). The analysis procedures are summarized in Fig. 1. A conjunction and contrast ALE meta-analysis was conducted in GingerALE (v3.0.2; Eickhoff et al. 2009) to identify domain-general (reading ∩ math) and domain-specific (reading > math and math > reading) areas. Region-of-interest (ROI) masks were created as 6mm-spheres around the peak coordinates of the significant domain-general and domain-specific clusters, which were used as seeds for subsequent MACM analyses (Eickhoff et al., 2011). Searches with each ROI mask were performed in Sleuth (v3.0.4), and the outputs were submitted to single-dataset analyses in GingerALE. For network modelling (Meier et al., 2021), mean ALE values from all ROIs were extracted from the MACM result for each seed in Mango. The p values representing covariance statistics were checked for significance with Bonferroni correction. Functional decoding was analyzed on the domain-general ROIs using the behavioral and paradigm analysis plug-ins in Mango (Lancaster et al., 2012).

The conjunction analysis and the contrast analyses of reading > math and math > reading yielded 4, 6 and 7 clusters, respectively (Fig. 2A). Hence, MACM analyses were performed with 17 seed ROIs. The conjunction of reading ∩ math identified 4 areas which were associated with working memory domain and EF paradigms as revealed by functional decoding analysis (z ≥ 3). Network modelling matrix showed that they were highly connected bidirectionally (Fig. 2B). Within the reading network, functional connectivity was found (1) from the reading-specific left middle temporal gyrus to the inferior frontal gyrus (IFG), (2) from reading-specific areas to domain-general areas, and (3) bidirectionally between the reading-specific left IFG (BA 9) and domain-general areas. For the math network, functional connectivity was found (1) bidirectionally between bilateral insulae and math-specific areas in the frontal and parietal lobes, (2) bidirectionally between math-specific and domain-general areas and (3) among math-specific areas.

This is the first study to examine functional connectivity in the domain-general and domain-specific networks between reading and math using a meta-analytic approach. The domain-general areas resembled the lateral frontoparietal network and the salience network (Uddin et al. 2019), which was behaviorally associated with working memory and functionally connected with reading- and math-specific areas. While the left IFG was the main hub connecting reading areas with domain-general areas in the reading network, highly bidirectional communications between math areas and domain-general areas were observed in the math network. Our results identified nodes and networks for future investigations on brain-behavior relationships to elucidate individual differences in reading and math skills.

Executive Function, Cognitive Control and Decision Making

Space, Time and Number Coding ²

Reading and Writing ¹

Activation (eg. BOLD task-fMRI)

fMRI Connectivity and Network Modeling

FUNCTIONAL MRI

Learning

Meta- Analysis

Other - Reading; Math; Executive Function; Working Memory; Development; Meta-analytic Connectivity Modelling