A Machine Learning Approach for Sex and Age Classification of Paediatric EEGs

Area of Science:

• Neuroscience
• Machine Learning
• Pediatrics

Background:

• Electroencephalography (EEG) is crucial for diagnosing childhood brain disorders.
• Visual EEG analysis can estimate a child's age by identifying maturational features.
• Determining a child's sex from EEG via visual inspection is not feasible.

Purpose of the Study:

• To investigate sex and age-related differences in EEGs of healthy children aged 6-10 years.
• To develop machine learning (ML) models for classifying sex and age from EEG data.
• To assess the potential of ML in age determination for distinguishing normal from delayed development.

Main Methods:

• Collected EEG data from 351 healthy male and female children (ages 6-10).
• Developed and applied ML algorithms to classify sex and age from EEG recordings.
• Evaluated model performance on a test set for accuracy in age and sex prediction.

Main Results:

• Achieved 66.67% accuracy in predicting child age within a 1-year error margin on the test set.
• The ML model performed poorly in estimating biological sex from EEG data.
• This preliminary study highlights ML's potential for age estimation in pediatric EEG.

Conclusions:

• Machine learning shows promise for aiding age determination in healthy children using EEG.
• Accurate age prediction from EEG may assist in identifying developmental delays.
• Current ML models struggle with accurate sex determination from pediatric EEG data.