Print  Close

Predictive validity of within and cross brain EEG connectivity indices of infant executive function

Poster No:

913 

Submission Type:

Abstract Submission 

Authors:

Lorena Santamaria1, Stanimira Georgieva1, Elizabeth Shephard2, Anthonieta Looman2, Daniel Fatori2, Guilherme Polanczyk2, Victoria Leong1,3

Institutions:

1Psychology, Nanyang Technological University, Singapore, Singapore, 2Department of Psychiatry, Faculdade de Medicina, Universidade de São Paulo, Sao Paulo, Brazil, 3Department of Pediatrics, Cambridge University, Cambridge, United Kingdom

First Author:

Lorena Santamaria  
Psychology, Nanyang Technological University
Singapore, Singapore

Co-Author(s):

Stanimira Georgieva  
Psychology, Nanyang Technological University
Singapore, Singapore
Elizabeth Shephard  
Department of Psychiatry, Faculdade de Medicina, Universidade de São Paulo
Sao Paulo, Brazil
Anthonieta Looman  
Department of Psychiatry, Faculdade de Medicina, Universidade de São Paulo
Sao Paulo, Brazil
Daniel Fatori  
Department of Psychiatry, Faculdade de Medicina, Universidade de São Paulo
Sao Paulo, Brazil
Guilherme Polanczyk  
Department of Psychiatry, Faculdade de Medicina, Universidade de São Paulo
Sao Paulo, Brazil
Victoria Leong  
Psychology, Nanyang Technological University|Department of Pediatrics, Cambridge University
Singapore, Singapore|Cambridge, United Kingdom
E-Poster

Introduction:

Executive functions (EFs) are core cognitive control skills that predict life success. These skills begin developing during the early years to allow children to sustain attention and resist distraction (inhibitory control), hold and manipulate information in mind (memory updating) and shift attention and strategies to adapt to changing demands (cognitive flexibility)(Fiske and Holmboe 2019; Lehto et al. 2003). EF development occurs within the context of positive social interactions and variations in the quality of parent-child interaction impact EF development. However, little is known about the intra- and interpersonal neural mechanisms (and their relative importance) in mediating influences of parent-child social interaction on early developing executive function skills. Here, we take adopt a computational machine learning approach to objectively contrast the feature importance of within and cross brain connectivity metrics on prediction of infant attention set-shifting performance (a precursor of cognitive flexibility). Importantly, we assess the reproducibility and generalisability of these neural indices by testing their predictive performance (1) within two different countries (each sampled separately) and (2) across countries (a stronger test of generalisation).

Methods:

A total of N=96 mother-infant dyads participated in this study across two countries, Brazil (N=57) and Singapore (N=39). Infants were aged, in days, 426±141.65 (SG) and 363±69.38 (BR) respectively. Infant and maternal brain activity were concurrently recorded via electroencephalography (EEG) whilst they performed an object play task to assess parental scaffolding of infant attention set-shifting (Tan and Leong 2023). Neural connectivity metrics (wPLI,(Vinck et al. 2011)) were computed using within-infant, within-mother and dyadic (mother-infant) EEG data. Identical pre-processing pipelines were used for all neural datasets (within and cross-brain). Graph theory-based metrics were calculated using three different thresholds (10%,15%,20%) to avoid possible bias(Bassett and Sporns 2017). Feature selection was performed using Mutual Information (MI) scoring to determine the best subset of "elite" neural features (pooling across within-infant, within-mother and dyadic indices) to enter into the predictive models. Understanding MI between two variables as a measure of the reduction in uncertainty for one variable given a known value of the other one. Two machine learning models (one linear and one non-linear) were implemented to predict infant shifting performance (2-class median split) using the elite neural indices(Nocedal and Wright 2006; Hastie et al. 2009). Model performance was evaluated using a leave-one-out cross-validation technique.

Results:

Overall, model classification performance achieved up to 86.2% accuracy within country, and 74.9% in the combined cross-country transfer scenario. Importantly, MI feature selection revealed that dyadic (mother-infant) metrics were the most important predictors of infant shifting in the cross-country transfer scenario, followed by maternal and infant brain metrics (see Fig 1A). This advantage for dyadic metrics was observed across both types of models (linear and non-linear), as well as across all three threshold values used. The higher MI total score of the dyadic feature indicates a stronger connection between these features and infant EF performance.
Supporting Image: MI_scores_together2.png
   ·Figure 1. Summary of the MI scores by member: dyad (red), infant (yellow) and mother (green). Three combinations were considered for the 2 cohorts (Brazil and Singapore): (A) both countries combined (
 

Conclusions:

Our results suggest that during early life, dyadic measures of parent-infant neural connectivity may provide robust and generalisable indices for the prediction of developing cognition, particularly emerging executive function skills. This empirical validation is an important first step toward developing reliable screening tools for early assessment of EF and its disorders.

Higher Cognitive Functions:

Executive Function, Cognitive Control and Decision Making 1

Lifespan Development:

Early life, Adolescence, Aging

Modeling and Analysis Methods:

Classification and Predictive Modeling
Connectivity (eg. functional, effective, structural) 2
EEG/MEG Modeling and Analysis

Keywords:

Development
Electroencephaolography (EEG)
Machine Learning
Modeling
PEDIATRIC
Social Interactions

1|2Indicates the priority used for review

Provide references using author date format

(Fiske et al. 2019; Lehto et al. 2003)
(Tan et al. 2023)
(Vinck et al. 2011)
(Bassett et al. 2017)
(Nocedal et al. 2006; Hastie et al. 2009)