A new artificial-intelligence system can estimate a cancer patient’s chance of survival from a single facial photograph. The tool was created by researchers at Mass General Brigham, a nonprofit academic health network in Massachusetts. It was trained on more than 40 million images covering a wide range of ages and ethnicities, allowing it to infer biological age and health status from visual cues.
The system produces two complementary outputs. FAHR-FaceAge calculates a person’s biological age, the age they appear to be, and also generates a survival estimate. FAHR-FaceSurvival bypasses the age step and predicts mortality risk directly from the picture. Each algorithm emphasizes different facial zones: FaceAge relies on creases from the nose to the mouth corners, the upper forehead, and the temples, while FaceSurvival focuses on the area beneath the eyes and the nasal bridge.
In a cohort of U.S. patients receiving radiotherapy, the model identified a clear survival gradient linked to apparent age. Individuals whose faces appeared five or more years older than their chronological age faced a 21 percent higher risk of death over an average four-year follow-up. Conversely, those who looked at least five years younger experienced a 17 percent reduction in mortality risk compared with peers whose facial age matched their calendar age.
The researchers interpret these findings as evidence that biologically older facial appearance reflects reduced physiological reserve, which can influence tolerance to aggressive cancer therapies. “As the population ages, physicians need better ways to estimate physical reserve and frailty before recommending intensive treatments such as radiation, surgery or systemic therapy,” said Dr Raymond Mak, a radiation oncologist at Mass General Brigham. The authors suggest the technology could eventually support personalized treatment planning.
Analysis of lifestyle variables revealed associations between certain habits and accelerated facial ageing as judged by the AI. Heavy alcohol consumption, illicit drug use and regular smoking were each linked to an older-looking face, suggesting that these behaviors may imprint measurable signs on the skin that correlate with health outcomes. Such correlations could enable clinicians to incorporate observable lifestyle effects into risk assessments without invasive testing.
Despite the promising results, the investigators stress that the approach remains experimental. Larger, prospective trials are required to confirm accuracy across diverse patient groups and to determine how best to integrate facial analysis into existing clinical workflows. Until such validation is achieved, the selfie-based tool should be regarded as an adjunctive metric rather than a definitive predictor of cancer prognosis.