Chinese researchers have developed a tool that uses AI to analyze images of an individual's face, tongue, and retina to determine their biological age. This technology provides insight into the health and condition of our cells, tissues, and organs and the risk of chronic disease.
A peek at a driver's license is enough to find out a person's chronological age. However, determining biological age accurately is difficult. Unlike chronological age, there is no universally agreed upon way to measure biological age, and it can be influenced by the environment, individual lifestyle choices, and genetics. For example, smokers can look years older than their actual age, while fitness enthusiasts can look much younger. And the difference is not superficial. If your biological age is greater than your chronological age, you may have a chronic illness or experience early cognitive decline. On the other hand, if you are biologically younger than your chronological age, you may have more momentum than others.
“Knowing your biological age is important because when you have time off, you can change your lifestyle and improve your health,” said Michael Snyder, a geneticist at Stanford University who was not involved in the study. “It is,” he says.
Early versions of aging clocks (models that measure biological age) look at DNA methylation patterns (chemical marks on DNA that control which genes are turned on and off) in different tissues as they change over time. I calculated this number by Other watches measured levels of various metabolic (blood sugar) and inflammatory protein markers that doctors often test during annual physicals. Recently, scientists designed a clock that confirms someone's biological age using his 3D image of the face, brain scans, and protein levels in the blood.
All of these track age-related changes, whether it's wrinkles in your skin or an increased likelihood of age-related diseases such as diabetes. However, aging is a complex process with myriad effects on multiple organ systems.
Kang Chan, a doctor and scientist at the Macau University of Science and Technology and co-author of the paper, said using one type of measurement to define biological age means that only the trunk of an elephant can be used. He says it's like trying to understand by touching it. PNAS In January.
Instead, Zhang and his collaborators are using an AI model that uses input from images of the face, tongue, and retina and spits out a corresponding age to create an “overall estimate” of biological age. ”I created an image. This methodology, similar to what powers ChatGPT, “examines vast amounts of data and discovers invisible connections that go beyond the ability to predict human age,” Zhang says.
“I was impressed by both the technical design of their deep learning experiments and the experimental design with the datasets they used. The results are very convincing,” said the author, who was not involved in the study. says James Cole, a neuroscientist at University College London.
Deep learning and biological age
Introduced in a 2017 Google paper, this AI technology (called Transformers) was first used to create programs that can mimic human language, such as ChatGPT. Unlike older AI models, Transformers process entire sequences of text at once instead of sequentially, making them more adept at detecting patterns and understanding context.
Soon, researchers translated their approach to image analysis, revolutionizing tasks in computer vision as well as natural language processing. Another innovation that some transformers, including the one used in this study, employ is analyzing images at different resolutions to extract both coarse and fine parts.
But transformers require a lot of data.
“Obviously, this isn't a language issue because there are a lot of languages out there, but for medical imaging it's much harder to find enough examples,” Cole says. His research applies AI techniques to brain scans to investigate relationships. Between aging and neurodegenerative diseases. He says it's great that the group was able to access tens of thousands of people for research.
As with many AI models, translating the model's results into human-understandable terms can be a problem.
“This model is subtle, focusing on pixel-level differences that we can't detect,” Zhang says. That said, their analysis shows that the center of the tongue (tongue image), the area around the eye (facial image), and the area with the highest concentration of blood vessels at the back of the eyeball (retinal image) are the areas that are associated with biological aging. An important reflection of.
face. tongue. retina.
To begin creating a tool that can determine biological age, researchers trained a model using images of the faces, tongues, and retinas of 11,223 people in northern China. Since they are healthy people, their biological age is assumed to be equal to their chronological age. This translates to 300 million variables, which is the second smallest number after ChatGPT4, which has 1 trillion parameters.
Prediction of chronological age as a proxy for healthy people's biological age is “more accurate within a year compared to other aging clocks” that use a single measurement, Zhang says .
Similar to the elephant story, information from each modality captures different aspects of aging.
For example, facial wrinkles suggest environmental factors such as sun exposure or pollution. On the other hand, thinning of the retina (part of the central nervous system) and damage to blood vessels reflect the health of the brain and circulatory system. Meanwhile, the shape of your tongue and its coating can provide clues about your microbiome as well as your gut health. As part of this study, study participants were followed for five years with regular health checkups, including blood tests, urine tests, lifestyle questionnaires, and physical exams.
With the biological aging clock in hand, Zhang's team tested the model on unhealthy people suffering from chronic diseases such as diabetes and heart disease, drawn from the same northern Chinese population used to develop the model. Tested. They also included people from other parts of southern China.
As expected, the biological age of healthy people was approximately the same as their chronological age. However, if someone has unhealthy habits such as smoking or a sedentary lifestyle, or has a chronic disease, their biological age tends to be older than their chronological age. This difference, called AgeDiff, ranges from an average of 3.16 years for people with chronic heart disease to an average of 5.43 years for smokers.
Consequences when biological age and chronological age are different
How being biologically older than your chronological age affects your risk of developing six common age-related diseases: chronic heart disease, chronic kidney disease, cardiovascular disease, diabetes, hypertension, and stroke To find out, researchers divided 11,223 people into four groups: AgeDiff from highest to lowest. People with high AgeDiff were more likely to develop one of these chronic diseases, and as AgeDiff increased, so did the risk.
Zhang was also interested in what AgeDiff could tell us beyond the risk of developing lifelong disease. when Disease may develop. In other words, will someone be diagnosed with diabetes this year or will he be diagnosed in five years? Knowing the timing “could actually be very helpful in designing interventions,” Zhang said. says.
What Zhang's team found is that AgeDiff is better at predicting when someone will get sick than traditional metrics such as blood sugar, BMI, and cholesterol. Combining AgeDiff with these other factors makes the predictions even more accurate. Not surprisingly, the researchers also found that abnormal values in health indicators such as BMI and blood pressure were associated with higher AgeDiff numbers.
The future of the body clock
Currently, Zhang and his team are using AgeDiff to identify people at high risk of developing chronic diseases and prescribing interventions for each of these patients. By targeting health indicators closely related to AgeDiff (for example, blood pressure and blood sugar levels), they hope to systematically delay the onset of the “diseases of aging.”
They are also improving the model by incorporating other variables such as DNA methylation and by incorporating experimental subjects from other ethnic groups.
Tools like AgeDiff could democratize health care and reduce the cost and effort of preventing disease before it spreads, Snyder said. To this end, Zhang's group is developing an iPhone attachment and associated app that can take the photos needed for the model, and hopes to have a working prototype by the end of the year.
Snyder, who studies how personalized medicine can reduce an individual's risk of developing chronic disease, likes this simple and accessible solution. “It's possible that anyone could do this very easily, without having to draw blood or all the tests that people are currently doing,” he says.
