# Predicting brain volumes from anthropometric and demographic features: insights from UK biobank neuroimaging data

## TL;DR

This study uses UK Biobank data to explore how brain volumes relate to body measurements and demographics, finding strong links with age and sex.

## Contribution

The study introduces a machine learning approach to uncover sex- and age-specific brain-body scaling relationships using large-scale neuroimaging data.

## Key findings

- Total intracranial volume (TIV) is strongly predicted by sex, with a correlation of 0.68.
- Age significantly affects total brain volume (TBV), gray matter volume (GMV), and cerebrospinal fluid (CSF) volume.
- Anthropometric measures like height and weight improve TIV and TBV predictions, but age and sex remain the primary factors.

## Abstract

Brain size measures are well-studied and often treated as a confound in volumetric neuroimaging analyses. Yet their relationship with body anthropometric measures and demographics remains underexplored. In this study, we examined those relationships alongside age- and sex-related differences in global brain volumes. Using brain magnetic resonance imaging (MRI) of healthy participants in the UK Biobank, we derived global measures of brain morphometry, including total intracranial volume (TIV), total brain volume (TBV), gray matter volume (GMV), white matter volume (WMV), and cerebrospinal fluid (CSF). We extracted these measures using the Computational Anatomy Toolbox (CAT) and FreeSurfer. Our analyses were structured in three approaches: across-sex analysis, sex-specific analysis, and impact of age analysis. Employing machine learning (ML), we found that TIV was strongly predicted by sex (across-sex \documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\:r=$$\end{document} 0.68), reflecting sex difference. On the other hand, TBV, GMV, WMV, and CSF were more sensitive to age, with higher prediction accuracy when age was included as a feature, highlighting age-related changes in the brain structure, such as fluid expansion. Sex-specific models showed reduced TIV prediction (\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\:r\:\approx\:$$\end{document} 0.25) but improved TBV accuracy (\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\:r\:\approx\:$$\end{document} 0.44), underscoring sex-specific body-brain relationships. Anthropometric measures, particularly seated height and weight, improved prediction of TIV and TBV, while waist and hip circumference showed negative associations, though their effects generally remained secondary to age and sex. These findings advance our understanding of brain-body scaling relationships and underscore the necessity of accounting for age and sex in neuroimaging studies of brain morphology.

The online version contains supplementary material available at 10.1007/s00429-025-03070-9.