An unbiased data-driven age-related structural brain parcellation for the identification of intrinsic brain volume changes over the adult lifespan

Abstract

This study aims to elucidate age-related intrinsic brain volume changes over the adult lifespan using an unbiased data-driven structural brain parcellation. Anatomical brain images from a cohort of 293 healthy volunteers ranging in age from 21 to 86 years were analyzed using independent component analysis (ICA). ICA-based parcellation identified 192 component images, of which 174 (90.6%) showed a significant negative correlation with age and with some components being more vulnerable to aging effects than others. Seven components demonstrated a convex slope with aging; 3 components had an inverted U-shaped trajectory, and 4 had a U-shaped trajectory. Linear combination of 86 components provided reliable prediction of chronological age with a mean absolute prediction error of approximately 7.2 years. Structural co-variation analysis showed strong interhemispheric, short-distance positive correlations and long-distance, inter-lobar negative correlations. Estimated network measures either exhibited a U- or an inverted U-shaped relationship with age, with the vertex occurring at approximately 45-50 years. Overall, these findings could contribute to our knowledge about healthy brain aging and could help provide a framework to distinguish the normal aging processes from that associated with age-related neurodegenerative diseases.

Keywords: Brain parcellation; Brain-age prediction; Healthy aging; Independent component analysis; Structural co-variation analysis.