Mass General Brigham develops AI tool predicting cancer survival from photos

US researchers have unveiled an artificial‑intelligence system that can gauge a cancer patient’s likely survival time from a single photograph of their face, the team reported in a study published this week in The Lancet Digital Health. The tool, dubbed “FaceSurvival,” was built by scientists at Mass General Brigham in Boston and tested on thousands of images from patients in the United States and the Netherlands. It is not yet approved for clinical use, but the researchers say it could eventually help doctors choose the right treatment intensity for frail patients.

The work comes as doctors grapple with an aging population and the need for faster, more objective ways to assess a patient’s physical reserve before recommending aggressive therapies such as surgery, radiation or chemotherapy. Traditional assessments rely on a battery of tests, physical exams and subjective judgment, which can vary between clinicians. By turning a routine headshot into a quantitative health metric, the researchers hope to add a scalable, repeatable layer to existing evaluations.

FaceSurvival was trained on millions of healthy faces and about 35,000 images of people diagnosed with various cancers, ranging from early‑stage tumours to disease that had spread to bone or brain. The algorithm focuses on subtle cues near the eyes, the bridge of the nose and the area under the eyes, which it links to overall frailty and expected survival. In the study, the system produced consistent estimates across a wide age span – from children to nonagenarians – and held up despite differences in lighting, pose, facial expression and even modest cosmetic surgery. The team also introduced a companion model called “FaceAge,” which predicts biological age by analysing nasolabial folds, marionette lines, forehead creases and cheek texture, noting that habits like heavy drinking, drug use and smoking accelerate the visual aging process.

The researchers caution that the technology is still in early stages and requires far larger, more ethnically diverse data sets before it can be trusted in a hospital setting. A patent application is pending, but the tool is intended to assist, not replace, physicians. Future work will focus on expanding the image library, validating predictions across different cancer types and integrating the model into clinical workflows. If successful, a simple, non‑invasive photo could become a rapid screen for frailty, helping clinicians tailor treatment plans and potentially improving outcomes for vulnerable patients.