Study maps local brain aging in postmenopausal women using deep neural networks

A neuroimaging study of 12,284 postmenopausal females using deep neural networks demonstrated that longer endogenous hormone exposure—indicated by later menopause onset and more births—predicts younger-appearing brains, highlighting the role of sex-specific endocrine pathways in neurological aging.

Researchers analyzed magnetic resonance imaging (MRI) and demographic data from 12,284 biological females in the UK Biobank, aged 46 to 82 years, using a 3D convolutional neural network with a modified V-net architecture.

The neural network was trained to calculate the local brain age gap (LBAG), which represents the difference between chronological age and biological brain age across 188 regions defined in the Destrieux atlas.

The study reported that longer reproductive span (beta values between -0.042 and -0.037 years per year), older natural menopause onset (beta values between -0.052 and -0.046 years per year), and a higher number of births (beta values between -0.230 and -0.190 per birth) were significantly correlated with younger biological brain ages.

In contrast, females with early menopause onset at or before age 46 exhibited a strong positive correlation with older biological brain ages (beta values between 0.338 and 0.392).

Exogenous hormone exposures, including history of postmenopausal hormone therapy and contraceptive use, did not exhibit a statistically significant correlation with local brain age, though both correlated positively with higher cognitive performance.

Women carry a disproportionately higher risk of developing Alzheimer's disease compared to men, making the identification of sex-specific biological aging trajectories critical for preventative neurogerontology.

The regional localized effects were strongest in Alzheimer's-vulnerable structures, particularly left temporal and frontal cortical regions, suggesting that prolonged endogenous estrogen exposure may convey regional neural resilience.

The divergence in results between endogenous exposure and exogenous hormone therapy highlights the complexity of timing and dosage in postmenopausal hormonal interventions, suggesting that early-life endocrine baselines may be more critical to long-term neuroprotection than postmenopausal treatments.

These findings bolster the clinical interest in upstream endocrine targets and estrogen-receptor modulation for preserving cognitive health, distinct from downstream protein-clearing modalities.

The validation of local brain age gaps (LBAG) via deep learning showcases a robust, reproducible biomarker that can be used by biotechnology and imaging companies to stratify neurodegenerative risk profiles in clinical trials.

Future longitudinal studies to verify whether regional LBAG alterations directly predict the velocity of cognitive decline, and if modified hormonal regimens can shift these trajectories in at-risk female cohorts.