An artificial-intelligence model developed by clinicians at Mass General Brigham can estimate a person's biological age and predict survival probabilities for cancer patients from a single facial photograph, according to reporting in The Guardian. The research was published on Monday in The Lancet Digital Health.
Two complementary tools built on the same model
The team created two complementary tools from the same AI backbone. FAHR-FaceAge first estimates biological age — a measure intended to reflect the body's physiological or cellular state rather than chronological years — and then derives implications for cancer patients' survival. FAHR-FaceSurvival instead attempts to predict survival directly from the face photo without first estimating biological age.
The tools focus on different facial regions: FAHR-FaceAge analyzes features such as folds extending from the sides of the nose to the mouth corners, the forehead area above the eye sockets, and the temples, while FAHR-FaceSurvival concentrates on the area beneath the eyes and around the nasal bridge.
Training data and study population
The system was trained on more than 40 million face images from people of varied ages and ethnic backgrounds, including about 35,000 images of people with cancer. The sample included predominantly US and Dutch participants aged between 50 and 60, but also ranged from children to individuals older than 90.
Findings: biological age estimates and mortality
The study found that AI-estimated biological age was associated with survival outcomes in cancer patients. Among patients in the United States treated with radiotherapy, those who appeared at least five years older than their chronological age had a 21% higher risk of death during a mean follow-up of four years compared with others. Conversely, individuals who appeared at least five years younger than their actual age had a 17% lower mortality risk.
The authors note that FAHR-FaceSurvival — which directly predicts survival from the photograph — also provided prognostic information, and that the two tools can offer complementary insights.
Lifestyle associations and robustness checks
The researchers observed links between certain lifestyle factors and facial aging as assessed by the AI: heavy alcohol consumption, drug use and smoking were associated with being judged older by the model.
They also tested potential confounders and reported that lighting conditions, body pose relative to the camera, facial expression and makeup did not substantially affect the tool's accuracy. Most cosmetic surgeries had only minor effects on results, although rhinoplasty produced a small observable change.
Clinical implications and need for further validation
The investigators suggest that biological age may reflect a patient’s physiological reserve and thus influence how well they tolerate intensive treatments. In future, the method could help clinicians weigh treatment options and serve as an inexpensive, accessible way to assess patient condition. The developers also propose potential use in estimating risks of mortality or complications after major surgeries.
However, they emphasize that broader clinical implementation requires additional validation studies, standardized testing and consideration of ethical and regulatory issues.
Conclusion
The Mass General Brigham AI system introduces a novel approach to deriving health- and prognosis-related signals from a single facial photograph. Early results are promising, but further independent validation is needed before the technology can be adopted widely in clinical practice.
