AI Tool Predicts Cancer Survival from a Patient's Facial Photo

Researchers at Mass General Brigham have developed an AI model, FAHR-FaceSurvival, that estimates cancer patients' mortality risk from a single facial photograph. Patients in the highest risk group had about a 50% death rate within 90 days, versus roughly 2% for the lowest risk group. The model, detailed in Lancet Digital Health, evolved from an earlier tool called FaceAge that predicts biological age. Its analysis focuses on areas below the eyes and along the nose, though the precise signals it detects remain unknown to scientists. Outside experts call the findings intriguing but caution that the tool must demonstrate it adds value beyond standard clinical assessments and must avoid encoding racial or other biases before practical use.