Mass General Brigham AI Predicts Cancer Survival from Facial Photos

Key points

• The AI tool, developed by experts at Mass General Brigham, was trained on over 40 million facial images to assess health metrics.
• Two complementary algorithms were created: FAHR-FaceAge estimates biological age, while FAHR-FaceSurvival directly predicts survival outcomes from facial images.
• FAHR-FaceAge analyzes specific facial regions, including nasolabial folds, forehead, and temples, whereas FAHR-FaceSurvival focuses on areas below the eyes and around the nasal bridge.
• In a study of US cancer patients undergoing radiotherapy, those appearing five or more years older than their chronological age had a 21% higher risk of death over a four-year period.
• Conversely, patients who appeared five or more years younger had a 17% lower risk of dying during the same follow-up period.
• The system demonstrated stability across various lighting conditions, poses, and most plastic surgery procedures, with only minor shifts observed after rhinoplasty.

Background

This development follows recent advancements in AI-driven medical diagnostics, such as tools that predict breast cancer risk from mammograms years before symptoms appear. While previous AI applications focused on analyzing medical scans or genomic data, this study represents a shift toward using ordinary facial photographs as a proxy for biological age and survival risk, building on broader trends in AI-enabled health assessment.

Why it matters

This technology could enable physicians to more accurately assess patient frailty and physical reserve before recommending intensive treatments like radiation or surgery. By providing a quantitative, scalable method for evaluating health status through simple photographs, the tool may help personalize treatment plans and improve outcomes for aging cancer patients, although it is intended to support rather than replace clinical judgment.

What to watch

Further research is needed to validate the tool's efficacy across diverse populations and clinical settings before it can be widely adopted in medical practice. Researchers aim to refine the model to ensure it remains robust against variations in facial appearance and to explore its application in predicting risks for other major surgeries and treatments.