Women's reproductive factors predict local brain aging profiles mapped using deep neural networks.

Abstract

Deep neural networks (DNNs) trained on magnetic resonance images can estimate global brain age (GBA), which reflects women's neurological disease risk. GBA gap (GBAG), the difference between GBA and chronological age (CA), quantifies excessive global aging; local BAG (LBAG) has not been examined despite allowing voxelwise resolution. Using a novel DNN architecture, we estimate LBAG for 12,284 UK Biobank females with chronological ages (CAs) ranging between 46 and 82 years (y) and quantify how it relates to cognition and women's health variables (CA at menopause, reproductive span, menopausal hormone therapy (HT), contraceptive use (CU), number of births). Women with longer reproductive spans (-0.042/y ≤ β ≤ -0.037/y, p < 0.001) had older CAs at menopause onset (-0.052/y ≤ β ≤ -0.046/y, p < 0.001), more births (-0.230 ≤ β ≤ -0.190 per birth, p < 0.001) and younger brains (more negative LBAGs, younger GBAs); a 1-unit increase in each of these variables reflects an LBAG change of β y. Left temporal lobe effects of CA at menopause onset are strongest (-0.0517 ≤ β ≤ -0.0510, p < 0.001). Cognitive scores are related to LBAGs negatively and most strongly in subcortical and right-hemisphere cortex (-0.021 ≤ β ≤ -0.017 per score unit, p < 0.01). In postmenopausal women, delayed regional brain aging is predicted by longer endogenous hormone exposure indexed by later menopause onset, longer reproductive span, and more births. This research highlights the complex role of women's health factors upon brain aging and related cognitive trajectories.