Brain Age Calculator

Seven numbers from a structural MRI report, turned into an estimate of how old your brain looks.

Scroll

How it works

A model reads your MRI numbers and guesses an age. The gap between that guess and your real age is the result.

The model is a Ridge regression fit on 150 subjects from the OASIS-2 research cohort. Enter your values below and the dial shows the predicted age, your real age, and the model's own margin of error. A gap inside that margin means nothing.

Typical error
±5.44yrs
Variance explained
22%R² 0.217
Reference scans
438sessions

Readout

405060708090100BRAIN AGE76.3+6.3 yrsYEARS · ±5.44 TYPICAL ERROR
Chronological
70yrs
Brain-age gap
+6.3yrs

These measurements read 6.3 years older than the age you entered. That is wider than the model's typical error of ±5.44 years, so the direction is probably real. Keep in mind the model only explains about a fifth of age variance.

The grey band on the dial is this model's typical error, ±5.44 years. A gap smaller than that band is inside the noise.

Your numbers

Sex

Biological sex as coded in the OASIS-2 cohort the model was fit on.

yrs

Your age in years. This is the mark the dial measures against.

typical 60–95
yrs

Total years of formal schooling. 12 is high school, 16 is a bachelor's degree.

typical 8–20

A rough 1 to 5 research scale, where 1 is lower and 5 is higher.

typical 1–5
mm³

Estimated total intracranial volume: the whole space inside the skull, brain and fluid together.

typical 1200–1900

Normalized whole-brain volume: the share of that space still filled with brain tissue. This one moves the dial most.

typical 0.6–0.82

Atlas scaling factor: a constant used to normalize for head size when the scan was registered.

typical 0.8–1.7

Where you land

Your reading against the reference cohort

Every dot is one scan session. The dashed diagonal is where predicted age equals real age. Use the legend to isolate a group.

You

Which inputs move the dial

← reads youngerreads older →
nWBV
−4.26
ASF
+3.14
Sex (male)
−1.54
Education
−0.39
Socioeconomic status
−0.19
eTIV
+0.01

Each bar is how many years the prediction moves when that input rises by one standard deviation, holding the rest fixed. Whole-brain volume does most of the work; education and socioeconomic status barely register.

How well it holds up

Typical error
±5.44yrs
Mean absolute error, subject-held-out cross-validation
Variance explained
22%R² 0.217
The other ~78% is not in these seven numbers
Demented group
+1.81yrs
Mean gap, reading older (n≈160)
Nondemented group
-1.29yrs
Mean gap, reading younger (n≈241)

Welch's t-test separates those two groups at t = 4.18, p < 0.0001. That separation holds at the group level. It is far too weak to say anything dependable about one person.

One result goes the wrong way

The Converted group (people who developed dementia during the study) has a mean gap of −2.34 years, the opposite direction from the Demented group. Most of their scans were taken before conversion, while they still looked cognitively normal, and that likely explains it. We report it as is.

Read before you believe any of this

  • This is an educational tool. It does not diagnose anything and it is not medical advice.
  • It was fit on 150 subjects from OASIS-2, a single research site. It has never been tested on anyone outside that population.
  • It explains about 22% of age variance and is typically off by 5.44 years. A gap inside that range is indistinguishable from noise.