Researchers at Mass General Brigham have developed artificial intelligence models that predict a cancer patient’s biological age and chances of survival from a single facial photograph, according to findings published in The Lancet Digital Health.
The study, led by Dr. Raymond Mak, demonstrates that digital facial analysis could serve as an additional tool for doctors to consider when evaluating a patient’s chances of survival and deciding which treatment approaches may work best. While chronological age simply counts years lived, biological age reflects how well the body has aged at the cellular and physical level. The AI models evaluate specific facial geometries to gauge overall health status in patients undergoing cancer treatment.
FAHR-Face Model Uses Specialized Algorithms for Aging
The research team initially trained a foundation artificial intelligence model named FAHR-Face on more than 40 million images capturing diverse ages and ethnic backgrounds. From this foundational architecture, the investigators derived two specialized algorithms: FAHR-FaceAge, which estimates biological aging, and FAHR-FaceSurvival, which estimates mortality risk directly from image data.
In validation analyses involving cancer patients, the algorithms isolated distinct anatomical zones to compute these metrics. The biological age algorithm assesses the temples, forehead, area above the eyes, and the nasolabial fold extending from the nose to the mouth corners. Conversely, the mortality risk algorithm focuses on the bridge of the nose and the region directly beneath the eyes.
Patients who appeared five or more years older than their chronological age exhibited a 21 percent higher risk of mortality. Among 122 patients flagged by the FAHR-FaceSurvival model as high risk, 52.5 percent died within 90 days, while only about 2 percent of the 975 patients in the lowest-risk group experienced the same outcome.
AI Measures Cellular Wear Versus Chronological Age
- Biological Age vs. Chronological Age: The AI measures cellular and physical wear and tear visible on the face rather than the number of years a person has been alive.
- Targeted Facial Zones: Different areas of the face are scanned to estimate aging versus predicting survival risks.
- Research Status: The tool is currently utilized for research and trials rather than routine diagnostic patient care.
Experts Call for Rigorous Clinical Evaluation of Tool
Despite these correlations, external researchers emphasize the need for rigorous clinical evaluation. Dr. Wilbur Lam, a research chair and professor of pediatrics and biomedical engineering at Emory University, noted that while the dataset is comprehensive, the tool still needs testing against doctors’ own judgement and across different hospitals and patient groups.
Concurrently, Mass General Brigham launched a public web portal where participants can upload selfies to test their own biological age via the FaceAge tool. Over 10,000 individuals have registered since March, contributing data to help refine the algorithms across broader cohorts. Researchers are also testing the technology in clinical trials, such as an evaluation at Brigham and Women’s Hospital involving patients ages 65 and older with early-stage lung cancer to determine if the tool can identify frail individuals requiring early nutritional or physical therapy interventions.
AI Scans Cannot Replace Clinical Diagnostic Evaluations
Patients and providers must recognize that experimental AI facial tools are not diagnostic devices and cannot replace physicians or existing clinical evaluations. Cosmetic alterations, including surgical procedures or injectable treatments like Botox, generally do not disrupt model performance—with only a small shift observed after rhinoplasty—yet visual scans cannot substitute for laboratory blood panels, imaging, or staging biopsies. Individuals undergoing cancer therapies should consult their primary oncologists regarding established prognostic markers rather than relying on experimental digital tools.
