This article is current as of February 2020. Details of how we calculate Brain Age may change in the future as we improve our tech!
When we think of the aging process, we naturally do so in terms of one number: calendar age. However, an MRI brain scan can add more detail to the picture. BrainKey’s Personal Brain Dashboard incorporates Brain Age, which provides an estimate of your age based solely on your brain scan. Comparing your Brain Age to your calendar age may provide insights into your aging process. As far as we know, Brain Key is currently the only service of any kind which provides a scientifically validated Brain Age estimate.
As you might expect, this comes with some caveats and disclaimers. Specifically — our products are meant to promote wellness, but we can’t claim to be actually diagnosing anything! Brain aging science is in its infancy, and it’s not yet known how best to apply it in the clinic. Here, we explain how we calculate Brain Age, and what your Brain Age might be telling you.
How do brains change as they age?
Representative brain images for 22-year-old (left) and 83-year-old (right) female scan subjects, with BrainKey segmentation. Note the deeper, wider grooves and thinner gray matter in the right-hand image. Scans sourced from publicly available datasets.
Generally speaking, as the brain ages, the outer layer of gray matter thins. The grooves (technically, the sulci) on its surface deepen and widen. While all brains undergo this general progression, different people age differently. Brain Age can be thought of as an attempt to sum up your progression in the aging process with a single number.
From brain scan to Brain Age
The Brain Age model we use was developed by Dr. James Cole, a London-based neuroimaging researcher and Brain Key advisor, and his colleagues. (Unless stated otherwise, references to Dr. Cole’s work here refer to this paper, which first described the model we use.) The Brain Age model converts your brain scan into a simplified image, boils the data down further with a technique called Principal Components Analysis (PCA), and then estimates age based on this simplified data.
To calculate your Brain Age, your scan is first converted into a simplified image showing the locations of gray matter, white matter, and cerebrospinal fluid (a fluid that surrounds the brain and fills its inner spaces, providing cushioning and taking away wastes). This is done by aligning the image to the MNI-152, a reference brain image created by averaging 152 high-quality MRI scans done at the Montreal Neurological Institute. This is a common first step in neuroimaging, and provides a consistent starting point for the next steps.
The preprocessed images are then boiled down to their most important features using a technique called Principal Components Analysis (PCA). The math behind PCA is complex, but it can be thought of as finding a representative set of brain images (the “eigenbrains”), then scoring each new image based on how similar it is to each eigenbrain. This can remove a lot of redundant detail; in the case of our Brain Age model, Dr. Cole and colleagues were able to use just 485 representative images to summarize his training data, meaning that each new image can be described with a list of just 485 scores.
These spooky “eigenfaces” resulted from a project at AT&T Laboratories — Cambridge which used PCA on a dataset of face photos.
Lastly, the 485 values extracted from PCA are combined using Gaussian Process Regression (GPR). This method often does a good job of extracting complex patterns from relatively small amounts of data (“only” 2001 brain scans were used to create Brain Age), making it a good choice for this task. After your brain scan is preprocessed, simplified using PCA, and fed into the final GPR model, a single value is generated: your Brain Age.
What might Brain Age mean?
More important than your Brain Age itself may be the difference between your Brain Age and calendar age — your “brain-predicted age difference” or “Brain-PAD”. Dr. Cole and colleagues determined the Brain-PAD of 669 individuals, all of whom were approximately 72 years old when scanned, and sorted them into equally sized groups of low, moderate, and high Brain-PAD. Because the scans were collected several years ago, the researchers were able to determine that individuals with low Brain-PAD were significantly more likely to still be alive 8 years post-scan than those with high Brain-PAD.
A larger proportion of individuals with low Brain-PAD survived to age 79 than those with high Brain-PAD. (The difference may appear bigger than it is, though! Note that 3 out of 4 of the high Brain-PAD group still survived.)
This image appears as Fig. 3b in “Brain age predicts mortality” by JH Cole et al. It is used under the terms of a Creative Commons v4.0 license.
Further, Dr. Cole and colleagues found that on average, individuals with higher Brain-PAD tended to perform slightly worse on a number of other tests of aging, such as tests of fluid cognitive performance, grip strength, and walking speed. Of course, this is a result from just one study; in particular, we don’t know much about what Brain Age may mean earlier in life. However, it does suggest that our Brain Ages may provide insight into other aspects of our health.
Influencing your Brain Age
If your Brain Age is higher than you’d like it to be, don’t panic! There’s some evidence that exercise, education, and meditation are each tied to a “younger-looking” brain. Uploading your brain scans to BrainKey and giving us permission to share them with researchers could speed up progress in this field, helping us understand what Brain Age means and how we can keep our brains healthy.
On a personal note, my Brain Age is about 15 years higher than my calendar age. I’ve been trying to get more exercise lately, and starting a brain imaging company has certainly been an educational experience — using BrainKey, I may be able to see if the gap between my Brain Age and calendar age has decreased next time I have a brain scan.
