Deep learning maps local brain aging in relation to cognition across human adulthood

Area of Science:

• Neuroimaging
• Artificial Intelligence
• Neurology

Background:

• Brain aging is a primary risk factor for Alzheimer's disease (AD), but global brain age (GBA) measures may obscure regional differences.
• Regional variations in brain aging could indicate areas vulnerable to cognitive decline and neurodegeneration.
• Existing methods may not fully capture the localized neuroanatomical changes associated with pathological aging.

Purpose of the Study:

• To develop and validate a deep learning model for estimating local brain age (LBA) at the voxel level.
• To map spatial patterns of brain aging in cognitively normal (CN) adults and individuals with mild cognitive impairment (MCI) and AD.
• To investigate the association between regional brain aging patterns and cognitive performance.

Main Methods:

• A deep learning architecture was trained on [Formula: see text]-weighted MRIs from 14,748 cognitively normal participants.
• The model estimated local brain age (LBA) at the voxel level, enabling detailed mapping of regional aging.
• LBAs were compared across different neurodegeneration stages (CN, MCI, AD) and correlated with cognitive performance.

Main Results:

• The model identified advanced aging in frontal and temporal lobes compared to parietal and occipital regions in CN adults.
• A pattern of progressively advanced frontotemporal aging was observed with increasing neurodegeneration severity (MCI to AD).
• Key AD-vulnerable brain regions showed significantly older LBAs in MCI and AD groups compared to CN adults, correlating with cognitive function.

Conclusions:

• Local brain age (LBA) mapping provides anatomically interpretable insights into regional brain aging patterns.
• This framework extends global brain age (GBA) measures by quantifying localized neuroanatomical decline.
• LBA analysis can enhance the characterization of typical and pathological brain aging, potentially aiding in early AD detection and understanding cognitive vulnerability.