AI uses a single face photo to assess patient health and survival

Researchers at Mass General Brigham built FAHR-FaceAge and FAHR-FaceSurvival on a model trained on over 40 million facial images to estimate biological age and mortality risk in cancer patients from a photograph. Among 122 high-risk patients, 52.5% died within 90 days versus about 2% in the lowest-risk group. The findings were published in The Lancet Digital Health.

• Model trained on more than 40 million facial images across ages and ethnicities
• 52.5% of high-risk patients died within 90 days versus about 2% at lowest risk
• Biological age 20+ years above chronological age linked to 21% 90-day mortality
• Heavy alcohol use, drug use and smoking tied to older estimated biological age