What Is the Brain Age Gap? How Brain Age Is Calculated

“Brain age” is an appealing idea, but the number that makes it scientifically useful is not the predicted age itself — it is the gap between that prediction and your real age. That single quantity, and the way it is computed, is what separates a serious neuroimaging biomarker from a party trick. This guide walks through the definition, the actual calculation, a methodological wrinkle that even many summaries miss, and how a tool can estimate the modifiable part of the gap without ever putting you in a scanner.

If you have not yet sorted out the three different things people call “brain age,” the hub guide on the three brain ages is the place to start. This page zooms in on the one that comes from neuroimaging.

§I.The formula, and what it means

The brain-age gap has a deliberately simple definition. Once a model has predicted an age for a brain, you subtract the person’s actual age:

Brain-Age Gap = Predicted Brain Age − Chronological Age

That is the whole equation. The result is a number of years, and its sign is what matters. A positive gap means the model thinks the brain looks older than the person’s real age — a signature of accelerated aging. A negative gap means the brain looks younger than its years, a pattern often read as resilience. The same quantity goes by several names in the literature: the brain-age gap, the brain age delta, or the brain-predicted age difference (brain-PAD).

A concrete example makes it tangible. Imagine a 60-year-old whose scan the model scores as a 55-year-old brain: their gap is −5, and their brain structure resembles that of someone five years younger. Now imagine a 50-year-old whose scan is scored as 57: their gap is +7, pointing to accelerated aging. Crucially, the gap is a deviation from the population norm — it only means anything relative to what is typical at that age, which is why building a good model starts with a large, healthy reference sample (Bethlehem et al., 2022).

§II.How is brain age calculated?

A machine-learning model is trained on brain scans from thousands of healthy people whose ages are known. It learns the normal pattern of how the brain changes with age, then predicts an age for a new scan from its structure. The gap is that prediction minus the person’s real age.

• Train on a large healthy sample so the model learns typical aging.
• Read the structure — cortical thickness, gray- and white-matter volumes, surface area, white-matter integrity.
• Predict an age for the new brain from those features.
• Subtract the real age to get the gap.

The original method, introduced in 2010, used a kernel-regression model on T1-weighted MRI scans and achieved a mean absolute error of about five years (Franke et al., 2010). The field has sharpened considerably since: deep-learning models that read raw images directly now correlate with true age at roughly r = 0.96 (Cole et al., 2017), and combining several imaging types tightens the estimate further (Cole, 2020). The largest models are now trained on more than fourteen thousand brains across the lifespan (Bashyam et al., 2020). In practice, the best current models predict chronological age to within about two and a half years (Franke & Gaser, 2019) — which is exactly why a credible result is always reported with a margin, never as a single false-precise figure.

§III.Why the gap is worth measuring

A brain-age gap would be a curiosity if it were random noise. It is not. Across large cohorts, a larger positive gap is associated with worse cognitive performance, higher rates of neurological and psychiatric conditions, faster future decline, and even higher mortality (Cole et al., 2018). That is what earns it the label biomarker: a single, interpretable number that summarizes something real about brain health and predicts meaningful outcomes. The value of the gap is not that it diagnoses a disease — it does not — but that it compresses a brain’s apparent aging into one figure that can be studied against everything else we know about a person (Cole & Franke, 2017).

§IV.The wrinkle most summaries skip: bias correction

Here is the part that separates careful brain-age work from the rest. Brain-age models carry a systematic bias: they tend to overestimate the age of younger brains and underestimate the age of older ones. The predicted age gets pulled toward the average age of the training sample — a statistical regression toward the mean. Left uncorrected, this makes the raw gap correlate with chronological age itself, which can manufacture associations that are really just age in disguise (de Lange & Cole, 2020).

Because of this, serious studies do not interpret a naive gap. They apply a statistical correction — typically regressing predicted age on real age and adjusting the estimate — before reading anything into the number (Smith et al., 2019). It is a small technical point with a large consequence: a brain-age gap is only trustworthy once it has been corrected for this bias. If you ever see a brain-age result reported without acknowledging it, that is a reason for caution.

§V.Two different calculations: the scan vs the questionnaire

Everything above describes the imaging brain-age gap, which needs an MRI and a trained model. But you have probably encountered “brain age” tools that ask questions instead of scanning you — including the one on this site. It is essential to understand that these compute something related but genuinely different, and conflating them is the single most common error in how brain age is discussed.

The Brain Age Index does not predict your brain’s age from a scan and subtract. It estimates the modifiable slice of the gap — the part driven by factors you can change — from your answers to evidence-based questions, weighted by the 2024 Lancet Commission’s risk factors (Livingston et al., 2024). In simplified form, its calculation is:

modifiable gap = Σ ( factor years × age-salience ) × 0.70

The 0.70 is a deliberate conservatism factor, applied because self-reported lifestyle and cardiometabolic factors explain only a limited share — at most about a fifth — of the variance in measured brain-age gaps. The rest is genetics, early-life development, and biology no questionnaire can see. The two calculations answer two different questions:

Neither is “more correct” — they are built for different jobs. The scan gives you the full biomarker but tells you little you can act on without a radiologist. The questionnaire deliberately isolates the part within your control, at the cost of seeing only that part. The honest move is to be clear about which one you are looking at.

§VI.How to read a brain-age gap honestly

Whichever calculation produced it, a brain-age gap is a deviation, not a diagnosis. The imaging gap comes with a margin of a few years, so small numbers are within the noise. The modifiable estimate is scaled down on purpose, because most of the true gap lies outside what self-report can capture. And a gap of either kind describes a pattern relative to a population, not a verdict about you: two people with the same gap can have genuinely different brains and different futures.

Read that way, the gap is genuinely useful — as a direction and a priority, not a prophecy. If a result, or a real concern about your memory or thinking, points to something clinical, treat it as a reason to talk to a qualified clinician rather than a conclusion from a number.

§VII.One number, two questions

The brain-age gap is powerful precisely because it is a single, interpretable figure: does this brain look older or younger than its years, and by how much? But that power depends entirely on knowing which gap you are looking at and how it was produced. The imaging biomarker, properly bias-corrected, gives you the full picture and needs a scanner. The modifiable-risk estimate gives you the actionable slice and needs only honest answers. Same idea, two calculations, two different questions — and once you can tell them apart, brain age finally makes sense.

@misc{lifebylogic_brain_age_gap_2026,
  title  = {What Is the Brain Age Gap? How Brain Age Is Calculated},
  author = {{LifeByLogic}},
  year   = {2026},
  url    = {https://lifebylogic.com/learn/brain-age-gap/}
}