Lifespan trajectories, metabolic associations, and links to fluid intelligence of brain temporal dynamics characterized by Multiscale Entropy

Abstract

Brain entropy (BEN) offers a unique perspective into brains functioning by characterizing the irregularity and randomness of brain activity. We used multiscale entropy (MSE) to estimate BEN from rs-fMRI at different timescales (from Scale 1 to Scale 5). Using a GMALSS framework, we showed that the aging trajectory of entropy differs at different timescales. Specifically, entropy increases with aging at Scale 1, the curve flattens at Scale 2, and the pattern finally reverses from Scale 3 onward, showing that entropy decreases with aging. Entropy at coarser scales (from Scale 2) showed positive correlations with measures of brain metabolism, including oxygen and glucose metabolism, whereas Scale 1 entropy displayed a trend for negative correlation. In addition, we explored the relationship between BEN and fluid intelligence (FI). Using moderated mediation models, we showed that brain entropy mediates the association between age and FI, and that the relationship between entropy and FI is itself moderated by age, indicating that entropy can either attenuate or exacerbate age-related differences in cognitive performance, depending on life stage. Finally, we used BEN to predict age. The best accuracy was achieved using the GPR (R2=0.80, MAE=7.25 years). Importantly, the difference between predicted and chronological age (brain age gap, BAG) was weakly associated with FI in an age-dependent manner, suggesting that lower entropy patterns relative to age-matched individuals may be associated with better cognitive performance during midlife, whereas higher entropy patterns may be associated with better cognition in older age.