AI tool estimates cancer patients' survival odds from one face photo
Mass General Brigham researchers built an AI that reads a single selfie to estimate biological age and predict cancer survival, as an aid to treatment decisions.
What happened: Researchers at Mass General Brigham built an AI system that estimates a person's biological age and predicts cancer survival odds from a single ordinary photograph of their face. Published in The Lancet Digital Health, the work produced two tools: FAHR-FaceAge, which estimates how old a patient looks biologically and links that to survival, and FAHR-FaceSurvival, which predicts survival risk straight from the photo without going through an age estimate. Both were built from a single underlying model trained on more than 40 million facial images spanning a wide range of ages and ethnicities.
Why it matters: Doctors treating cancer patients, especially older ones, often have to judge by eye whether someone can physically tolerate radiation, surgery or chemotherapy. That judgment is subjective and inconsistent. A quick, low-cost photo-based score could give physicians a more standardized read on a patient's physical reserve before recommending aggressive treatment. The researchers frame it as a decision-support aid, not a replacement for a doctor's own assessment, but it points toward biometric photos becoming a routine data point in cancer care.
How it works, plainly: The two tools look at different parts of the face. FAHR-FaceAge focuses on nasolabial creases, the forehead above the eyes, and the temples to judge how old someone looks. FAHR-FaceSurvival instead scans the under-eye area and nasal bridge to estimate survival risk directly. In a group of radiotherapy patients followed for about four years, those who looked five or more years older than their actual age had a 21% higher risk of dying; those who looked five or more years younger had a 17% lower risk. Heavy drinking, drug use and smoking were linked to looking biologically older. The scores held up reasonably well across different lighting, poses, expressions and makeup, with only a small shift after rhinoplasty.
The rollout: This is a research result, not a bedside tool yet. The study's own authors say more validation is needed before the system is used to guide real treatment decisions, and they stress it's meant to support, not override, a physician's judgment. Open questions include how it performs across different cameras, image quality and populations not well represented in training data, and how hospitals would handle consent and privacy around using patients' facial photos as medical data.