A Manifold Framework for Interpretable Brain Age Estimation and Aging Trajectory Mapping

Brain aging involves complex structural changes that challenge traditional analytical methods and hinder personalized assessment of brain health. Integrating regional brain alterations into a unified, interpretable representation is difficult due to the high dimensionality of neuroimaging data. Here, we projected regional brain volumes from the Human Connectome Project Aging Dataset onto a low-dimensional manifold that reflects underlying neuroanatomical constraints. We then built a transparent framework on the manifold to estimate brain age and identify key regional drivers of the estimate. By analyzing local neighborhoods on the manifold, we identified distinct structural aging trajectories, including pronounced frontal atrophy occurring predominantly in males. This approach provides a biologically interpretable means of characterizing individual brain-aging patterns, reveals heterogeneous aging pathways, and supports more personalized assessments and insights into the aging process.