What the researchers built
Researchers at Mass General Brigham in Boston have developed artificial intelligence models that, they say, can estimate a cancer patient's biological age and likelihood of survival from a single photograph of the face. The findings were published Monday in The Lancet Digital Health, according to CNN.
The work started with a foundation model called FAHR-Face, trained on more than 40 million facial images covering a range of ages and ethnic backgrounds. The team then fine-tuned it into two task-specific tools. FAHR-FaceAge estimates biological age from one photo. FAHR-FaceSurvival predicts a cancer patient's chances of survival by combining facial features with other clinical information.
Biological age differs from chronological age, which counts years lived. It is meant to reflect how well the body has aged at the cellular and physical level.
What the data showed
In testing the survival model, 122 cancer patients were classified as high risk of death. More than half of them, 52.5%, died within 90 days of the assessment. Among 975 patients in the lowest-risk group, only about 2% died in that period.
The age model showed a similar pattern. Of 1,060 patients whose estimated biological age was more than 20 years above their chronological age, 21% died within 90 days. Among 4,195 patients whose biological age was at least 10 years younger than their actual age, about 6% did.
The researchers also reported that cosmetic changes such as surgery, injectables like Botox or makeup did not affect how the models performed.
Which parts of the face matter
Dr. Raymond Mak, a radiation oncologist at Mass General Brigham Cancer Institute and a study author, said the two models focus on different areas. The age algorithm looks at the temples, the forehead above the eyes and the nasolabial fold, the line running from the nose to the corner of the mouth. The mortality model looks at the bridge of the nose and the area under the eyes.
Mak noted that these regions echo a description by the ancient Greek physician Hippocrates, who wrote of "the nose sharp, the eyes sunken, the temples fallen in" as signs of muscle wasting and mortality risk. He said the algorithms were trained only on face photographs and on whether someone died or on their age, not on any text. "To think that our algorithms might be replicating that wisdom of Hippocrates from millennia ago was mind-blowing to us," he said.
Caution from outside experts
The study authors stress the tools are not meant to replace physicians or existing clinical evaluations. They describe them as an emerging aid that could someday inform a doctor's assessment or treatment decisions. Mak said that identifying accelerated aging or higher mortality scores early might help flag patients who could benefit from earlier testing. The study involved only cancer patients, though he said further research could examine whether similar results hold for otherwise healthy people.
Dr. Wilbur Lam of Emory University, who was not involved, called it "an extremely compelling study" with associations that "appear to be robust," but said its clinical utility "will need to be demonstrated prospectively." He also said it remains a question whether the tool works accurately for specific subpopulations, including people with particular facial features, skin color or age groups.
Miroslaw Bober of the University of Surrey, speaking in a statement distributed by the UK-based Science Media Centre, said the study offers "convincing evidence" of a statistical link between facial appearance and survival in patients receiving radiotherapy. But he said the AI estimates age from appearance rather than directly measuring biological ageing, and that the study did not show the predictions add useful information beyond doctors' assessments of general health and fitness. He called it "a promising proof of concept, but it is not ready for clinical use."
