Brain age

Brain age is a machine-learning estimate of how old your brain appears structurally, compared to a chronological-age reference. It's derived from a single T1-weighted MRI scan run through a model trained on the brains of thousands of healthy adults.

The model has learned what brains tend to look like at 25, at 40, at 70 — the cortical thickness, white matter integrity, the volumes of structures like the hippocampus and the ventricles. Given a new scan, the model returns a single number: the estimated age of that brain.

Why it's useful

Chronological age is a count of time. Brain age is a measure of how that time has actually been spent — biology, lifestyle, illness, protection. Two people born on the same day can have brain ages that differ by 10 years or more. The discrepancy turns out to carry real information about cognitive ageing, recovery from neurological injury, and risk of neurodegeneration.

The single most useful derived metric is the brain age gap — the difference between predicted and chronological age.

Where it's used

• Cognitive ageing research — to characterise trajectories of brain change with age
• Stroke recovery — accelerated brain ageing after stroke predicts worse motor outcomes
• Drug trials — as a sensitive imaging endpoint for treatments aimed at slowing brain ageing
• Personal curiosity — increasingly available outside research as the methods mature

How it's computed

Most current methods use convolutional neural networks (e.g., SFCN) trained on large healthy reference cohorts like the UK Biobank. The input is a preprocessed T1-weighted scan; the output is a single scalar.

A robust pipeline includes bias-field correction, registration to a standard template (MNI152), skull-stripping, intensity normalisation, and resampling to the model's expected input shape.

Read more

• Two ages, one body — a non-technical overview
• How stroke accelerates brain ageing — recent work in stroke recovery