AI Tool Predicts Cancer Survival by Analyzing a Patient's Selfie

A single snapshot of a cancer patient's face may someday help forecast how long they'll live. Researchers at Mass General Brigham say they've built an AI model, FAHR-FaceSurvival, that estimates mortality risk using only a facial photograph, according to a study in Lancet Digital Health. Among patients whose scores landed in the top 10% for risk, about half died within 90 days; in the lowest 10%, roughly 2% did, the Washington Post reports.

The system grew out of an earlier tool, FaceAge, which guesses a person's "biological age"; that system collects different information about a patient's health, per the Guardian. Attention maps suggest the new AI focuses on areas below the eyes and along the nose, though scientists don't yet know what precise signals it's reading. Outside experts call the work intriguing, per the Post, but stress that it must prove it adds information beyond what doctors can already see—and that it avoids encoding racial or other biases.