Decoding brain age predictions from sleep electroencephalography across infancy to adolescence

Area of Science:

• Neuroscience
• Developmental Biology
• Biomedical Engineering

Background:

• Childhood sleep electroencephalography (EEG) patterns reflect brain maturation.
• Age-aligned EEG provides insights into neurodevelopment.
• Objective markers of neurodevelopmental health are needed.

Purpose of the Study:

• To predict chronological age from overnight EEG in children using supervised neural networks.
• To derive a Functional Brain Age (FBA) across different sleep stages.
• To assess FBA accuracy and reliability as a marker of neurodevelopmental health.

Main Methods:

• Utilized supervised neural networks to analyze overnight EEG data from 814 children.
• Calculated FBA across wake, NREM (N1-N3), and REM sleep stages.
• Evaluated prediction accuracy, influencing factors (EEG channels, sleep segments, data quality), and explainability.

Main Results:

• Achieved an overall Mean Absolute Error (MAE) of 0.96 years for FBA prediction.
• Prediction accuracy varied developmentally (infancy MAE: 0.78 years, childhood MAE: 0.87 years, adolescence MAE: 1.55 years).
• FBA accuracy was highest during N2, N3, and REM sleep stages, correlating with known developmental EEG changes.

Conclusions:

• Functional Brain Age derived from sleep EEG is a reliable, non-invasive tool for tracking childhood brain maturation.
• FBA can serve as an objective marker for neurodevelopmental health.
• Scalable EEG-based tools can enhance monitoring of child brain development.