Researchers led by Mass General Brigham in Boston built an artificial intelligence system called FAHR-Face and reported it in The Lancet Digital Health, published on 19 October. The system was trained on more than 40 million facial photographs to pull health indicators from facial features. The team split it into two tools: FAHR-FaceAge, which estimates biological age, meaning how old the body looks, and FAHR-FaceSurvival, which predicts survival from a photo directly. Each tool reads different parts of the face. FAHR-FaceAge relies on the folds running from the sides of the nose to the mouth, the forehead above the eye sockets and the temples, while FAHR-FaceSurvival focuses on the area below the eyes and around the bridge of the nose. The authors say the cancer estimates were checked against real patient data.
The survival link was tested in 38,183 cancer patients treated with radiotherapy in the United States. Those who looked at least five years older than their actual age had a 21% higher risk of dying during follow-up, which averaged four years, than the rest of the group. Patients who looked five or more years younger had a 17% lower risk. The age model was first checked on photographs it had not seen before, and it was off by 5.1 years on average. For 4,892 radiotherapy patients, the average gap between estimated and real age grew to almost seven years, which the authors read as a sign of accelerated biological ageing. In the separate survival model, built on 34,363 photographs of cancer patients, the highest-risk quartile had more than three times the mortality of the lowest-risk quartile after adjusting for age, sex, cancer type, treatment intent and radiotherapy technique.
Data from the article text
The researchers also found that certain lifestyle factors were linked to looking older. Heavy alcohol use, drug use and smoking were each associated with an older-looking face. Smoking showed the strongest pattern: current smokers looked older than former smokers, who in turn looked older than people who had never smoked. These are associations found in the study's photo analysis, and they sit alongside the survival results rather than replacing them.
Hugo Aerts, a co-senior author of the study, said the model learns survival risk directly from patient photographs.
“To our knowledge, this is the first demonstration that an AI model can directly predict cancer survival from a simple photograph.”
Raymond Mak, associate professor of radiation oncology at Harvard Medical School, said the team found consistent results across differences in pose, lighting, expression and makeup in real-world photos. He described the aim of the work in these terms.
“Our goal is to make face photograph-based health assessment more quantitative, consistent and scalable so it can support personalised treatment decisions rather than replace physician judgment.”
The model remains preliminary, with a long way to go before it shows clinical use. Members of the public can take part in the research by uploading a photo to the study's website and receiving their own readout.
Pollar is free and carries no ads in the text. Its members keep it going.
Support PollarDiscoveries · Boston
Researchers at Mass General Brigham trained FAHR-Face on more than 40 million facial photographs. In a Lancet Digital Health study, cancer patients who looked older than their actual age had a higher risk of dying after radiotherapy.
AI text based on reports by EL MUNDO, Bloomberg Business, The Guardian and other publishers. Published .