A Low-dimensional Manifold Representation of the Human Brain Aging Continuum

Abstract

The structural changes accompanying brain aging exhibit complex, multifaceted patterns that challenge traditional analytical approaches and impede accurate assessment of individual brain health. While previous studies have documented these age-related changes, integrating findings across multiple brain structures to assess individual brain age status has proven difficult due to the high-dimensional nature of neuroimaging data. Using structural MRI data from the Human Connectome Project Aging Dataset, we demonstrate that age-related regional volume changes can be mapped onto a low-dimensional manifold that reflects underlying neuroanatomical constraints. This manifold representation provides a transparent framework for both assessing an individual\'s overall brain aging status and identifying specific regional changes that drive this assessment. By examining data within neighborhoods on the manifold, we identified distinct structural changes within anatomically homogeneous subgroups, including pronounced frontal atrophy that was observed primarily in male participants. These local patterns can be aggregated to create a comprehensive picture of how an individual brain ages, providing new insights into the heterogeneous nature of brain aging.