Abstract
The functional repertoire of the human brain changes dramatically throughout the developmental trajectories of early life and even all the way throughout the adult lifespan into older age. Capturing this arc is important to understand healthy brain ageing, and conversely, how injury and diseased states can lead to accelerated brain ageing. Regression modelling using lifespan imaging data can reliably predict an individual's brain age based on expected arcs of ageing. One feature of brain function that is important in this respect, and understudied to date, is neural oscillations-the rhythmic fluctuations of brain activity that index neural cell assemblies and their functioning, as well as coordinating information flow around networks. Here, we analysed resting-state magnetoencephalography (MEG) recordings from 367 healthy participants aged 18 to 83, using two distinct statistical approaches to link neural oscillations and functional coupling with that of healthy ageing. Spectral power and leakage-corrected amplitude envelope correlations were calculated for each canonical frequency band from delta through gamma ranges. Spatially and spectrally consistent associations between healthy ageing and neurophysiological features were found across the applied methods, showing differential effects on neural oscillations, with decreasing amplitude of low frequencies throughout the adult lifespan, and increasing high-frequency amplitude. Functional connectivity within and between resting-state brain networks mediated by alpha coupling generally decreased throughout adulthood and increased in the beta band. Predictive modelling of brain age via regression showed an age-dependent prediction bias, resulting in overestimating the age of younger people (<40 years old) and underestimating the age of older individuals. These findings evidence strong age-related neurophysiological changes in oscillatory activity and functional networks of the brain as measured by resting-state MEG and that cortical oscillations are moderately reliable markers for predictive modelling. For researchers in the field of predictive brain age modelling with neurophysiological data, we recommend attention is paid to predictive biases for younger and older age ranges and consider using specific models for different age brackets. Nevertheless, these results suggest brain age prediction from MEG data can be used to model arcs of ageing throughout the adult lifespan and predict accelerated ageing in pathological brain states.