A simple selfie could hold surprising clues about your overall health and longevity, especially for people with cancer.
Researchers at Mass General Brigham in Boston have developed artificial intelligence models that they say can predict a cancer patient’s biological age and chances of survival using nothing more than a simple photograph of their face.
The research, published Monday in the journal The Lancet Digital Health, suggests that facial analysis could become an additional tool for doctors to consider when assessing a patient’s chances of survival and deciding which treatment approaches may work best for their cancer patients.
“I was quite surprised to find that we could detect signs of mortality risk from a simple facial photograph,” said study author Dr. Raymond Mak, a radiation oncologist at Mass General Brigham Cancer Institute and faculty member in the Artificial Intelligence in Medicine program at Mass General Brigham.
The AI models – trained on more than 40 million facial images – specifically evaluate different aspects of the face to assess the general health and biological age of a cancer patient. Unlike chronological age, which measures the number of years a person has lived, biological age reflects the body’s aging on a cellular and physical level.
“The age estimation algorithm looks at the area around the temples, the forehead area, above the eyes, and then what we call the nasolabial fold, the line that connects the nose to the corner of the mouth,” Mak said of the parts of the face that AI models consider when assessing biological age.
“In contrast, the mortality risk prediction algorithm looks at a completely different part of the face,” Mak said. “It’s all about looking at the bridge of the nose and the area under the eyes.”
Strangely, the same regions of the face that the AI model evaluates to predict biological age and survival are the same parts that the Greek physician and philosopher Hippocrates referred to in the Book of Prognostics, written around 400 BC, when he described “pointed noses, sunken eyes, drooping temples” as signs of muscle wasting and risk of mortality.
“To think that our algorithms could replicate the wisdom of Hippocrates from millennia ago was mind-blowing to us,” Mak said.
The AI algorithms had no knowledge of the ancient texts of Hippocrates, because “our algorithms are trained only on photographs of faces and then whether someone is dead or not, or their age,” Mak said.
“This does not take into account the texts, knowledge or wisdom collected on the Internet,” he added. “So to hear Hippocrates echoing for millennia in our algorithms was very strange.”
Mak and his colleagues initially built a robust AI model called Foundation Artificial Intelligence for Health Recognition-Face (FAHR-Face) that was trained on more than 40 million facial images spanning diverse ages and ethnicities.
The model was then refined and used as a basis for building two other models capable of performing specific tasks: the FAHR-FaceAge model estimates a person’s biological age based on a single photo, and the FAHR-FaceSurvival model predicts a cancer patient’s chances of survival by combining facial features with other clinical information.
With the FAHR-FaceSurvival model, researchers found that among 122 cancer patients indicated by the model among those at high risk of death, more than half – 52.5% – died within 90 days of assessment, as expected. By comparison, only about 2% of the 975 patients in the highest-risk FAHR-FaceSurvival group died within 90 days.
In evaluating the FAHR-FaceAge model, researchers found that 21% of 1,060 patients whose biological age was more than 20 years older than their chronological age died within 90 days. In comparison, approximately 6% of the 4,195 patients had a biological age at least 10 years younger than their chronological age. died within 90 days.
Researchers found that cosmetic modifications to the face – such as surgery, injectables like Botox or makeup – did not affect the performance of AI models.
The study authors emphasize that their AI models are not intended to replace doctors or existing clinical assessments. Instead, they describe the technology as an emerging tool that could one day help inform a doctor’s assessment of a patient’s health status or decision-making about treatment options.
“If we can detect accelerated aging or higher mortality scores earlier, we may be able to screen for them or identify patients who might benefit from additional testing for serious illnesses earlier,” Mak said.
And although the new study was conducted only in cancer patients, Mak said additional research could reveal whether similar results could emerge in otherwise healthy people who might be interested in their longevity.
The use of AI technology in the study is fascinating but still needs to be evaluated clinically, said Dr. Wilbur Lam, research chair and professor of pediatrics and biomedical engineering at Emory University.
“This is an extremely compelling study that arises from a large data set and its associations between facial images and biological age as well as cancer survival risk appear robust. Its true clinical utility, however, will need to be demonstrated prospectively,” Lam, who was not involved in the study, said in an email.
“Additionally, as the authors themselves allude to, whether this tool can be used accurately for specific subpopulations remains a question,” he said. “For example, the strength of effects on people with specific facial characteristics, skin color and age groups will also need to be tested prospectively so that the technology can be implemented clinically.”
The study provides “compelling evidence” of a statistical link between facial appearance and survival in cancer patients receiving radiotherapy, Miroslaw Bober, professor of video processing and artificial intelligence at the University of Surrey, said in a statement released by the UK-based Science Media Centre.
But “predicting an outcome is not the same as explaining it. Despite references to ‘biological age,’ AI estimates age from facial appearance rather than directly measuring an underlying biological aging process,” said Bober, who was not involved in the study.
“The authors associate the facial regions that the AI relies on with the ‘dying look’ described by doctors. However, the study did not test whether the AI recognizes something that a doctor does not recognize,” Bober said.
“These predictions have not been shown to add clinically useful information beyond doctors’ assessments of a patient’s general health and fitness, which were not measured in the study,” he said. “This is a promising proof of concept, but it is not ready for clinical use. The tool still needs to be tested against doctors’ own judgment and in different hospitals and patient groups.”
Gn Health
