EEG Neurosubtyping Infants Predicts Language Development

Latent profile analysis of infants at risk for ASD (n=144)
📄 Read Full Article in Journal of Neural Transmission

Summary

Key takeaways from our research on infant brain patterns and language development.
🧠

Three Brain Pattern Groups

We found three patterns of brain activity in 6-month-old babies using brain recordings.

📈

Predictive of Language

These early brain patterns predict language skills through age 3.

👂

Left-Side Advantage

Infants with stronger left-hemisphere activity showed better language outcomes.

🔍

Early Identification

This approach could help identify infants who might benefit from early language interventions.

Audience Level:

Understanding Infant Brain Development

We studied 144 infants to identify brain activity patterns that might predict language development. By mapping these patterns early, we can better understand how the infant brain processes language and potentially identify children who may benefit from early interventions.

Key Research Questions

  1. Can we identify distinct brain activity patterns in 6-month-old infants?
  2. Do these patterns predict language development outcomes?
  3. How do these patterns relate to autism spectrum characteristics?

Research Approach

1

Data Collection

EEG recordings from 144 infants (6-7 months)

2

Brain Pattern Analysis

Identifying neural signatures using advanced algorithms

3

Group Classification

Sorting infants into three distinct brain pattern groups

4

Language Assessment

Following development from 6 to 36 months

🔊

Language Development

Language is one of our most important human abilities. For children with autism, speech and language difficulties are often one of the first concerns raised by parents. The more severe these communication challenges are, the earlier a child typically receives an autism diagnosis.

🧠

Brain Differences

We still don't fully understand why language delays happen in the brain. This is because autism presents differently in each child, with varying patterns in learning ability, language skills, and sensory responses making it challenging to study.

📊

Our New Approach

Instead of comparing children with and without autism, we look at brain patterns to identify natural groupings. This "brain subtyping" approach has proven effective at uncovering meaningful differences that can help guide personalized interventions.

📈

EEG Brain Waves

We use EEG to safely measure brain activity in infants. By analyzing different types of brain waves and connections between brain regions, we can identify patterns that may predict which children will need extra support with language development.

Research Impact

Early brain measures can predict language development trajectories, contributing to evidence that brain-based stratification approaches can help decode developmental heterogeneity and guide personalized early interventions.

Who We Studied

Participant characteristics and measured EEG biomarkers (n=144).
Participant Demographics
Characteristic n %
Diagnostic Status
ASD 22 15.0
No ASD 114 79.0
ASD Likelihood
Typical 75 52.1
Elevated 69 47.9
Site
London 67 46.3
Washington 77 53.5
Sex
Male 76 52.8
Female 68 47.2
EEG Measures & Associated Brain Regions
EEG features and their brain locations
Key EEG Features: 1) Frontal Gamma Power, 2) Connectivity Auditory Network, 3) Connectivity Language Network, 4) Connectivity Speech Network, 5) Spectral Power Lateralization

Gender Considerations

Analysis revealed significant gender differences in language development, with females showing higher receptive language scores across groups (β=-0.06, p=0.029). This highlights the importance of considering sex as a variable in early neurodevelopmental research.

How We Analyzed Brain Activity

Our approach to analyzing EEG patterns and tracking language development over time.
Project Methodology
Project methodology overview
A) Data was collected from a cohort of 6-7 month-old infants (n=144), comprising 96 with elevated likelihood of autism spectrum disorder and 48 with typical development. B) EEG measures were extracted from key brain regions, primarily those associated with language processes. C) EEG features were pre-processed and analyzed to be used for stratification methods. D) Both hierarchical clustering and latent profile analysis was used to derived three distinct EEG subtypes. Above is a dendrogram showing the three subgroups identified. E) Results from the latent profile analysis and hierarchical clustering were analyzed for each group. Analyzes showed connectivity in the left hemisphere, lateralization of gamma power and connectivity in the auditory network to be important for deriving groups. F) Longitudinal language development trajectories (Mullen Scales of Early Learning scores) from 6 to 36 months across the three identified EEG subtypes.
📊

Latent Profile Analysis (LPA)

A probabilistic approach that identifies latent groups based on multivariate normal distributions of observed variables. LPA was used to classify infants into distinct groups based on their EEG features.

🧩

Hierarchical Clustering (HC)

A deterministic method that builds nested clusters by merging or splitting groups based on distance metrics. HC provided a complementary approach to confirm the group structure identified through LPA.

What We Discovered

Three distinct brain activity patterns identified through multiple analytical approaches.

AGroup A Profile

  • High connectivity in left language areas
  • Leftward brain lateralization
  • Strongest language outcomes

BGroup B Profile

  • High connectivity in auditory network
  • No brain lateralization
  • Lower language scores

CGroup C Profile

  • Overall low connectivity
  • No brain lateralization
  • Intermediate outcomes

🧩Autism Spectrum Findings

  • No ASD-specific phenotype: Children later diagnosed with autism were distributed across all three groups (χ²=1.15, p=0.561 for LPA; p=0.974 for HC)
  • Clinical relevance: Findings support individualized assessment approaches rather than "one-size-fits-all" interventions for children on the spectrum

Brain Regions of Interest

Interactive 3D visualizations of brain regions involved in language processing for each identified group.
Group A Brain Connectivity
High connectivity in left language areas with leftward lateralization.
Group B Brain Connectivity
High connectivity in the auditory network with no brain lateralization.
Group C Brain Connectivity
Overall low connectivity across networks with no brain lateralization.

The Three Brain Pattern Groups

Our analyses revealed three distinct groups using both Latent Profile Analysis ? and Hierarchical Clustering ? methods.
Sankey Diagram
Class Sizes
Class Relationships
Flow diagram showing relationship between LPA (left) and HC (right) classes.
Class Distribution
Class sizes visualization
Distribution of participants across identified classes.

Feature Distribution Analysis

Distribution of EEG features across the identified groups, showing how neurophysiological markers ? differ between groups.
EEG Distributions
Feature Discriminative Power
EEG Feature Distributions
Distribution plots for EEG features
Distribution of each EEG measure across clusters, comparing LPA and HC classification approaches.
Feature Discriminative Power
Feature discriminative power
F-statistics showing which EEG features most strongly differentiate between groups.

Brain Activity Profiles by Group

Visualization of EEG profiles for each identified group, showing patterns of neurophysiological functioning.
Radar Plots
Parallel Coordinates
Profile Heatmap
LPA Group Profiles (Radar)
Radar plot of EEG profiles for LPA groups
Radar plot showing normalized EEG profiles for LPA groups (A, B, C).
HC Group Profiles (Radar)
Radar plot of EEG profiles for HC groups
Radar plot showing normalized EEG profiles for HC groups (a, b, c).
LPA Profiles (Parallel Coordinates)
Parallel coordinate plot for LPA groups
Parallel coordinate plot showing individual trajectories and means for LPA groups.
HC Profiles (Parallel Coordinates)
Parallel coordinate plot for HC groups
Parallel coordinate plot showing individual trajectories and means for HC groups.
Combined Profile Heatmap
Heatmap of EEG profiles across LPA and HC groups
Heatmap with hierarchical clustering of EEG profiles across LPA and HC groups.

Visual Mapping of Brain Patterns

Dimensionality reduction showing separation of groups in EEG feature space using UMAP ?.
3D UMAP Projection
2D UMAP Projection
2D UMAP Projection
2D UMAP showing cluster separation
2D UMAP projection showing how subjects cluster in the EEG feature space.
3D UMAP Projection (Interactive)
Interactive 3D UMAP showing EEG feature space colored by LPA and HC groups.

How Language Develops Over Time

Longitudinal analysis of language development by group, showing how different EEG-based groups develop language abilities from 6 to 36 months.
Trajectories Graph
Expressive Language Model
Language Trajectories
Language trajectories by group
Developmental trajectories of language skills across EEG groups over time.
Expressive Language Model Results
Fixed effects table for the expressive language linear model.
Receptive Language Model Results
Fixed effects table for the receptive language linear model.

Language Abilities at 3 Years

Analysis of language abilities at 36 months, showing differences in outcomes between EEG-based groups.
Language Scores at 36 Months
Boxplots of language scores at 36 months by group
Boxplots comparing expressive and receptive language scores at 36 months by group, with statistical significance annotations.

Key Finding

Groups identified through EEG profiles show statistically significant differences in language outcomes at 36 months, with Group A/a showing the most favorable outcomes.

Key Insights & Implications

Important discoveries about early neural markers and their implications for language development.

Early Predictive Power

Brain patterns at 6 months can predict language development trajectories, suggesting early intervention possibilities.

Autism Independence

No differences in autism diagnosis across groups, indicating these brain patterns reflect general language development mechanisms rather than autism-specific traits.

Left-Hemisphere Advantage

Left-hemisphere bias in brain activity during infancy appears beneficial for language development, consistent with previous research.

Nonverbal IQ Impact

Nonverbal cognitive abilities at 6 months strongly predicted later language skills, highlighting the interconnected nature of cognitive development.

Connecting Brain Patterns to Language Skills

Understanding the connection between early neural markers and language development.
🧠

Neural Substrates

The neural foundations for language development are present from early infancy, with the speech network and left-hemisphere lateralization playing crucial roles.

📊

Stratification Approach

Multivariate stratification approaches prove valuable in understanding developmental heterogeneity and may guide personalized interventions.

🔬

Research Impact

This research demonstrates the potential of using early brain measures to predict language development trajectories, contributing to evidence that brain-based stratification approaches can help decode developmental heterogeneity and guide personalized early interventions.