Machine learning algorithms for age prediction based on linear and non-linear parameters of electroencephalogram data

Abstract

Gaining insights into cognitive and behavioral changes during childhood and adolescence requires a fundamental understanding of the developmental trajectory of the human brain. This research aimed to predict the age of children using linear and non-linear measures of baseline electroencephalogram (EEG) data. EEG is a method that records the electrical activity of the brain, providing valuable insights into its functioning. Participants were 182 children between 7 to 20 years old. Peak alpha and entropy were correlated with age. Various machine learning models were implemented, with Decision Trees yielding the best results. The Decision Trees model achieved strong correlation between predicted and actual age. The study demonstrated the stability of age prediction error over time, suggesting individual brain maturational levels. The findings highlight the potential of EEG data for accurate age prediction, providing insights into brain maturation patterns. This research contributes to tracking neurodevelopment and understanding brain function across age groups, including typically developing children.