Electroencephalography estimates brain age in infants with high precision: Leveraging advanced machine learning in...

Area of Science:

• Developmental Neuroscience
• Computational Neuroscience
• Clinical Neurophysiology

Background:

• Early detection of atypical neurodevelopment is critical for timely intervention.
• Current prognostic tools for infant development are often delayed, relying on cognitive and behavioral assessments after the first year.
• Electroencephalography (EEG) offers a promising avenue for early, non-invasive assessment of brain maturation.

Purpose of the Study:

• To develop and validate a machine learning (ML) methodology, specifically deep learning (DL) networks, for quantifying the brain age gap (BAG) using infant EEG.
• To assess the efficacy of the BAG tool in differentiating typically developing infants from those with abnormal brain growth (macrocephaly).
• To explore the relationship between the EEG-based BAG and later cognitive outcomes in infants with macrocephaly.

Main Methods:

• A cross-sectional study utilizing EEG data from 219 typically developing infants (3-14 months).
• Development of conventional ML and novel DL models to predict biological age from EEG.
• Validation of the DL-based BAG tool in a cohort of infants with macrocephaly, correlating BAG with neurodevelopmental assessments (ABAS-II, WPSSI-IV).

Main Results:

• Deep learning networks significantly outperformed conventional ML models, achieving a mean absolute error of 1 month in predicting biological age (r = 0.82).
• The EEG-based BAG successfully identified group-level maturational delays in infants with macrocephaly compared to typically developing infants (p=0.009).
• In macrocephaly, a negative correlation was observed between BAG and later adaptive functioning (ABAS-II, p=0.04) and information processing speed (WPSSI-IV, p=0.006).

Conclusions:

• The developed EEG-based brain age gap (BAG) score is a reliable, non-invasive marker for assessing infant brain maturation.
• Deep learning models effectively capture complex EEG patterns indicative of developmental trajectories, including the role of alpha rhythm.
• The BAG tool has significant potential for early detection of developmental deviations and informing clinical practice in developmental neuroscience.