The Washington Post
WASHINGTON — Imagine that an insurance underwriter comes to your house and, along with noting your weight and blood pressure, snaps a photo of your face. And that those wrinkles, mottled spots and saggy parts, when fed into a computer, could estimate how long you will live.
Facial recognition technology, long used to search for criminals and to guess how a missing child might look as an adult, may soon become personal. A group of scientists is working on a system that would analyze an individual’s prospects based on how his or her face has aged.
“We know in the field of aging that some people tend to senesce, or grow older, more rapidly than others, and some more slowly,” said Jay Olshansky, a biodemographer at the University of Illinois at Chicago who came up with the idea. “And we also know that the children of people who senesce more slowly tend to live longer than other people.”
The research is still in its early stages, but the idea of using facial recognition technology has prompted interest from insurance company executives who see potential for using it in determining premiums, Olshansky said. There’s also a potential benefit for individuals: The technology might prod them to change their health habits before it’s too late.
The technology involves using a computer to scan a photograph of a face for signs of aging. Factoring in the subject’s race, gender, education level and smoking history — all known to affect longevity prospects — it would analyze each section of cheek, eye, brow, mouth and jowl looking for shading variations that signal lines, dark spots, drooping and other age-related changes that might indicate how the person is doing compared with others of the same age and background.
As the United States skews increasingly older, research into extending life span and, in particular, increasing the number of healthy years is a boom topic for public and private entities.
Google last fall announced Calico, a new enterprise focusing on aging and associated diseases, for which it has been recruiting top scientists; it has not revealed details of its plans or how much it is investing. Another organization, Human Longevity Inc., headed by the well-known genomics researcher Craig Venter, launched this spring with plans to build a database of human DNA sequencing to tackle diseases of aging; it raised $70 million in an initial round of funding.
And the National Institutes of Health recently launched an unprecedented collaborative initiative across 20 of its 27 specialized institutes to address aging and longevity. National Institute on Aging director Richard Hodes said the NIH would also like to work on the topic with some of the emerging organizations.
The economic and social implications could be staggering. Not only will living to 100 become more common one day, longevity experts say, but the quality of life in the final decades might also be drastically improved, reducing the burdens imposed by an aging population.
Increasing life expectancy by 2.2 years by slowing aging would save $7.1 trillion in disability and entitlement programs over 50 years, according to a paper in Health Affairs co-authored by Olshansky, who is also a research associate at the University of Chicago’s Center on Aging.
Longevity scientists say the key to extending healthy life lies in focusing on aging itself rather than on aging-related diseases. Even minor progress in slowing the aging process would be more groundbreaking than major progress that tackles just one illness, they say.
In fact, drugs already in use for some age-related diseases may turn out to work because they are delaying aging overall.
“We may be at the beginning of a time when drugs approved for diabetes or macular degeneration are actually working because they are delaying the onset of aging,” said Dan Perry, founder of the Alliance for Aging Research, a Washington-based advocacy group.
And while it is not yet clear whether humans will one day live 150 years, as some have predicted, scientists are optimistic that the number of years of healthy life — or “health span” — of humans can be significantly increased and the infirmities associated with aging reduced.
“Aging is not such a deep part of our biology that it can’t be changed,” said Steven Austad, chair of the biology department at the University of Alabama at Birmingham. “All this stuff seemed like science fiction a few years ago, but now we have it, at least in mice.”
The idea for the facial recognition project came to Olshansky a couple of years ago during dinner with an insurance underwriter. “He was complaining that he had a very short time to assess people’s survival prospects” and that the methods used to do it were too blunt, Olshansky said.
Olshansky, whose work includes exploring the limits to human longevity, slowing aging and studying health and public-policy implications of individual and population aging, knew that people who live longer generally look younger than other people of their age. He wondered whether that knowledge could translate into something more scientific.
He contacted Karl Ricanek, a professor of computer science at the University of North Carolina at Wilmington, who has worked on facial recognition technology for the National Security Agency, the CIA, and the FBI; along with a biostatistician and other computer scientists, they developed a program to analyze photographs of faces.
They have launched a website inviting anyone in the world to submit a photo. The database they are developing, called Face My Age, is expected to deliver increasingly more accurate assessments and predictions as more people participate. The researchers are hoping for large numbers of people — at least 10,000 or 20,000, but preferably more — to submit photos and basic biographical information in exchange for feedback on how quickly they are aging and what this means for their longevity prospects. The person in the photo cannot smile or have makeup on, and must reveal if he or she has had plastic surgery.
The technique is more personalized than the current approach to face aging.
“The technology that is out there utilizes group norms, so they can artificially age you,” Ricanek said. “But … the lines they paint on your face are actually the same as the lines they paint on my face, whereas the ones we’re using are individual.”
Initially the site will give users only one number — their apparent age — but as it becomes more refined, it should be able to assign perceived ages to different parts of the face, Olshansky said.
“Imagine taking your iPhone and snapping a selfie and putting it into our website and discovering that your eyes are that of a 50-year-old, your lips are that of a 70-year-old, your cheeks are that of a 50-year-old,” he said.
The algorithms work differently for people of different genders and ethnic groups, Ricanek said. For example, the skin of lighter-complected individuals, which has less melanin, tends to age more as a result of sun exposure than the skin of people with darker complexions. Women’s faces tend to age more quickly than men’s because of different distributions of fat and blood vessels.
It won’t be clear how well the technology works until enough participants die and the researchers can see how good their estimates were. But the project recently got a boost when it gained access to several thousand photos taken years ago of people, some of whom have subsequently died; knowing the date of death for so many will allow the website to start providing users with even more reliable life span estimates in the next 12 to 18 months, Olshansky said.
If successful, it could be used not only by insurance companies but also by health advocates, financial institutions and other scientists.
The concept is intriguing — if it works, said Nir Barzilai, director of the Institute for Aging Research at the Albert Einstein College of Medicine in New York. But he said it is not clear whether skin appearance alone can reveal deeper signs of aging.
“You really want to see if the skin biomarker is associated with other disease,” he said.
Barzilai, who works with centenarians, said he plans to submit photos of some of his subjects, ages 60 to 116, to the database.
James Kirkland, director of the Kogod Center on Aging at the Mayo Clinic in Rochester, Minnesota, said that more important than estimating a person’s life span would be predicting his or her functional state, which Ricanek and Olshansky’s database will not do. But like many discoveries that end up contributing to science in unexpected ways, “it could be part of a pipeline that eventually results in something,” he said.
Ethical and practical concerns may also arise, said Leonard Fleck, a professor of philosophy and medical ethics at Michigan State University.
Even if it can predict life span, the analysis might not be able to predict a person’s need for long-term care, he said. And it could open the door for discrimination.
“If at age 40 if there were something about your face saying you’re not likely to make it past 60, an employer could say, ‘Oh, I’m not willing to promote you to some position of importance because it’s not likely to be a good investment,’ ” Fleck said.
And people who look younger than their years do not always last long, said Mark Collins, president of the California-based Glenn Foundation, which funds aging research. “Sometimes people who look very healthy drop dead in the middle of the track, while others who look crinkled are still running at age 80,” he said.
Olshansky conceded that even if face aging is found to correlate with longevity, there will be outliers who don’t fit the general pattern.
“The longest-lived person in the world smoked for 100 years,” he said, adding that U.S. presidents, too, tend to be outliers, aging visibly faster in office but generally living longer than average.
However, he said, for the most part a face is a window onto a person’s overall health.
“The face picks up a lot of risk factors for health, such as tobacco smoking (wrinkles around the mouth); excessive alcohol consumption (larger nose); and excessive exposure to the sun (early brown spots and wrinkling) as well as stress,” he said.
At the very least, learning the results of one’s face-age analysis may nudge participants to try to extend their healthy life spans by adopting good habits.
“If someone came to you and said that your life expectancy, for example, is five years from now, you would think pretty hard and long about what’s going on in your life,” Ricanek said. “It can make us wake up and change some of the things that we’re doing — maybe we’re stressing out too much about our job; maybe we need to make different lifestyle decisions. I would like to shake people up.”