how old you? The answer is written on your face.

Inspired by the centuries-old Chinese practice of medical practitioners to predict a person’s health by “reading” their faces, Han and her team, computational biologists at Peking University, have been working with Jidong’s approximately 5,000 We created a clock by analyzing 3D facial images of human inhabitants. China. Researchers created two AI-derived clocks. One predicts chronological age and the other predicts biological age. These facial aging clocks track the changes our face undergoes over time, such as drooping eyes, widening nose, drooping chin, and increasing distance between nose and mouth. Also, certain facial features are known to match certain diseases. For example, systemic inflammation manifests itself in sagging skin.

Han’s research has the potential to upend preventive medicine, according to Andre Esteva, founder and CEO of a medical AI startup in Los Altos, California. your lifestyle. The tool also allows doctors to track and manage the care of patients undergoing troublesome treatments known to lead to premature aging, such as chemotherapy. It may also be useful in aging research.

“We have received a lot of requests from companies who want it. [our tool] To assess the effectiveness of supplements and anti-aging drugs,” says Han.

The AI ​​model needs examples where it already knows the correct answer, or “ground truth,” and learns how to look at it in new data (e.g. faces paired with age of subjects). So a potential problem arose. There is no gold standard for biological age. “The concept of biological age is a collective term for all multisystem phenomena that occur with age,” explains Christopher Bell, who studies the relationship between age and chronic disease at the University of London. From shortening telomeres (the caps that keep chromosomes from breaking down) to shrinking mitochondria to a weakened immune system, it’s hard for him to pick just one marker of aging.

of The first clock to measure aging It was based on changes in the pattern of methyl groups. This is a chemical tag added to DNA that turns genes on and off. This DNA methylation machinery that regulates gene activity deteriorates over time. The pattern of this deterioration (which regions of the genome are affected) tells us how quickly cells and tissues are aging.Other watches that rate age are based on Protein distribution in blood Also Number of times a stem cell divides.

face aging clock

In 2016, Han decided to use a different scale. It’s your perceived age, or how old other people look.she was inspired by 2009 survey Volunteers assessed the age and health of the twins by photograph, followed by a medical evaluation of the subjects seven years later. found that they were frail, cognitively impaired, and more likely to die. Clearly, Han thought that biological age, as surrogated by physical appearance, was strongly correlated with health.Her group already has her 3D image of her face Research published in 2015found that facial features can predict chronological age in a statistical model developed using 300 subjects in Beijing.

At the time, when she was investigating markers, or signatures, in the blood and ascertaining their relationship with age, she worked at the Institute of Computational Biology, a partner of the Chinese Academy of Sciences and the Max Planck Society. I noticed that there is a 3D facial esthetic that I was doing. imager. She said they had this imager, so she took it with her when she collected blood samples and used it to compare what the face reveals about aging with what the blood markers reveal. She thought it should be used.

By 2016, she had access to a larger cohort, about 5,000 people in Jidong, and more powerful AI technology. Her AI’s data processing power allowed Han to replicate the human perception of how biological age appears on a person’s face. To minimize variability from a single observation, in Han’s new study, each subject’s biological age was assessed independently by her five volunteers to provide a “ground truth” for training the AI. functioned as

The AI ​​engine, which learned and improved from these human observations, was surprisingly accurate. On average, both chronological and cognitive age models deviate their predictions from actual age by about 3 years. In Han’s research, he called people who look three years older than their birth certificate date premature aging. Slow Age, on the other hand, made him look three years younger.

This difference between predicted age and actual age can be correlated with various health parameters, Han said. According to Han, these outliers tell us the most about how we age.

To explore the links between physical appearance, age, and underlying factors that affect health, Han took blood samples and studied the habits of study subjects. For example, smoking, snoring, and high levels of total cholesterol in the blood are often characteristic of premature aging, whereas yogurt consumption, regular meal times, and increased bone density are common features of late aging. It was a characteristic feature.

The middle age from the 40s to the early 50s seems to be the time when the difference between the first age and the slow age becomes more pronounced. Some people look very old at 40, and some look very young at 55. According to Han, this variability means that interventions to adopt healthier habits can make a big difference in this window of opportunity.

Han and her collaborators explored the underlying molecular mechanisms behind aging and also built additional models to match facial features of rapidly aging humans via an AI-derived clock. . For example, they developed a clock that identifies active genes in the blood of individuals who appear older and matches them to facial features of the same individual.

For overlapping predictions, researchers can see which genes (and thus molecular pathways) are turned on in older faces compared to younger faces. For all watches, “accelerated aging is strongly associated with infection and inflammation,” it appears on the face as narrowing of the forehead due to sagging skin. , full cheeks and bulging bags under the eyes.

Application of technology

Aging research has become mainstream, but one of the problems in the field has always been: Biological age is neither precisely known nor objectively defined quantitatively. But according to Ruibao Ren, a medical scientist at the International Center for Aging and Cancer at Hainan Medical University in China, Han’s study, like ongoing work in other labs, addresses this information gap. We have made great strides in narrowing it down.

“Just being able to measure aging is a big step forward, and Dr. Han’s technology, which is already being used at the Hainan hospital where he works, will play an important role in aging research,” said Ren. says Mr. Pharmaceutical companies can use this tool to evaluate drugs that extend lifespan or slow aging, for example. As an oncologist, Ren hopes that the facial clock can be used to diagnose cancer patients when they are in their early stages. Cancer is strongly correlated with age, so if someone is showing signs of accelerated aging, doctors can screen more aggressively. is the cause,” says Ren.

Physicians can also incorporate biological age as one of the panels of tests included in annual physical exams, such as cholesterol tests and blood pressure measurements. “How are you managing your health?” Ren says.

Also, unlike other watches that require blood or tissue samples and expensive analyzes to be implemented, the face clock is relatively economical and non-invasive, requiring the patient to sit in front of the 3D imaging camera for only one minute. is enough.

Our goal

Bell also sees great potential for further study of the aging clock, especially as the dataset grows. We looked at the sites, and the latest clocks account for over 900,000. But the real power of the aging clock lies in its potential to reveal new insights into the aging process, says Bell.

For example, how aberrant DNA methylation sign cancer and aging. By understanding both scenarios, “we may be able to identify new therapies to treat these diseases,” he says.

However, he advises caution in interpreting the results of different aging clocks. “At the population level, these clocks capture some aspects of aging.” At the individual level, however, there is still work to be done. “If you measure the age of someone’s particular epigenetic clock, you can’t be sure how much that individual’s tracking and changes over time represent reality.”

As for Han, her team continues to explore other clocks, such as transcriptomic clocks that reflect DNA damage in blood and single-cell clocks based on information from a single cell. The aim is to aggregate different measurements into a composite clock, each of which indicates different aspects of aging. We are also continuing to improve our facial aging clock and will soon have a version applicable to all ethnic groups.

What is Han’s ultimate purpose? “Well, I tell her colleagues to try to make them live at least five years longer and look at least five years younger,” she laughs.