Normative EEG growth curves of infant visual neurodevelopment generalize across global cohorts
Poster No:
1035
Submission Type:
Abstract Submission
Authors:
Chris Camp1, Emma Margolis2, Ana Sobrino2, Kirsten Donald3, Guilherme Polanczyk4, Daniel Fatori4, Takao Hensch5, Charles Nelson5, Elizabeth Shephard4, Dustin Scheinost6, Laurel Gabard-Durnam2
Institutions:
1Yale University, New Haven, CT, 2Northeastern University, Boston, MA, 3University of Capetown, Capetown, Capetown, 4University of Sao Paulo, Sao Paulo, Brazil, 5Boston Children's Hospital, Boston, MA, 6Yale School of Medicine, New Haven, CT
First Author:
Co-Author(s):
Introduction:
Normative growth curves, such as those charting height and weight during development, map an individual's measurements onto a population norm. Centile scores representing an individual's position along this axis can then be investigated in relation to various outcomes. This approach has been successfully applied to characterize structural neuroimaging features across the lifespan. However, there is a need to understand normative functional brain development and further test these growth curves for generalizability and stability in multinational longitudinal data. We sought to characterize the rapid visual neurodevelopment that occurs in infants using the visual-evoked potential (VEP), a robust visual response that can be measured with electroencephalography (EEG).
Methods:
We leverage generalized additive models for location, scale, and shape (GAMLSS) to create longitudinal normative growth curves of VEP morphology with 1,374 VEPs collected across 802 infants (57 to 579 days old) from South Africa, Brazil, and the United States. Models were cross-validated using a leave-one-site-out approach. Model fits were evaluated by Pearson r correlation and root mean square error of the actual and predicted VEP feature values. We combined centile scores from the six VEP features to create a composite deviation score, which we used to investigate the relationship between VEP neurodevelopment and cognition.
Results:
Site-specific models were cross-validated and showed excellent fits to other sites' samples, demonstrating functional growth curves generalize across contexts robustly [Fig. 1A]. The cross-validation average correlation ranged from r = .68 (N2 Latency, 95% CI: 0.63–0.74) to r = .99 (N1 Amplitude, 95% CI: 0.99–0.99), and the average room mean square error was 11.85. Individual-level VEP centile scores were significantly related to Global Scales for Early Development (GSED) cognitive development scores (r(90) = .26, p = .012)[Fig. 2]. Individual centile scores were generally stable across development, despite substantial changes to the distribution of VEP feature values themselves [Fig. 1B]. We also replicated previous work identifying a positive relationship between fetal alcohol exposure and latency of the P1 component of the VEP using centile scores (r(130) = .45 controlling for age at EEG, p < .001), further underscoring the relevance of the VEP to infant health and development.
·Figure 1
·Figure 2
Conclusions:
The robustness of VEP growth curves across highly distinct global contexts and the ability to predict later cognition from derived centiles demonstrates the utility of functional neurodevelopmental growth charts. These growth curves are an asset to global public health efforts and early identification efforts to improve intervention and support healthy brain development globally. To aid further development and application of these reference curves, we will make the normative growth charts freely available within the HAPPE EEG processing software to extract centile scores for VEP data in new cohorts.
Lifespan Development:
Normal Brain Development: Fetus to Adolescence 1
Modeling and Analysis Methods:
EEG/MEG Modeling and Analysis 2
Methods Development
Neuroanatomy, Physiology, Metabolism and Neurotransmission:
Normal Development
Perception, Attention and Motor Behavior:
Perception: Visual
Keywords:
Cognition
Development
Electroencephaolography (EEG)
Modeling
NORMAL HUMAN
Statistical Methods
Vision
Other - Normative Modeling
1|2Indicates the priority used for review