Develop a sleep electroencephalogram (EEG) age norm from 0 to 100 years old based on large-scale datasets.

Establishing an age norm in normal participants provides a reference when reading an individual’s EEG. Many studies have proposed sleep EEG age norms in different portions of the lifespan, but most are derived from small datasets, therefore having limited generality in clinical practice. 

The dataset is based on 6,365 polysomnograms from healthy subjects aged 11 days to 94 years from three large-scale datasets: the Massachusetts General Hospital dataset (N=5,132; age range: 4 to 94 years old), the Collaborative Home Infant Monitoring Evaluation [N=407; age range: 11 days to 8 months old), and the Gozal Pediatric dataset (N=826; age range: 0 to 13 years old). We took the common four EEG channels (C3-A2, C4-A1, O1-A2, and O2-A1) and computed the spectrum of each 30 second EEG epoch using the multitaper spectrum estimation method. For each sleep stage and EEG channel, we fitted a parametric model which takes age as input and outputs a spectrum. Data from 95 to 100 years old was extrapolated. 

Delta (0.5-4 Hz) and theta (4-8 Hz) power bands are highest from birth to 10 years of age, particularly in stage N3 sleep. Alpha (8-12 Hz) power is low until 15 years of age, particularly in REM and stage N1 sleep, increasing thereafter. Spindle (12-14 Hz) power in stages N2 and N3 sleep is highest from 10 to 70 years old, peaking in middle age, and being minimal outside this age range. Beta (13-20 Hz) power stays low throughout all sleep stages and is the lowest of all band powers. Conversely, delta power has the highest band power throughout the spectrogram. 

Comprehensive age norms of sleep EEG were established using large-scale datasets providing a reference for clinical practice.