Researchers at Mass General Brigham in Boston developed an artificial intelligence model that predicts a cancer patient’s biological age and chances of survival using a single facial photograph, according to a study published on Monday in The Lancet Digital Health. The findings, reported by gulfcoastnewsnow.com, suggest that automated facial analysis could eventually assist doctors in evaluating patient health and determining treatment approaches.
Training the FAHR-Face Models on 40 Million Images
The research team built a core foundation model called Foundation Artificial Intelligence for Health Recognition-Face (FAHR-Face), which was trained on more than 40 million facial images spanning diverse ages and ethnic backgrounds. From that foundation, the team created two specialized tools. The FAHR-FaceAge model estimates a patient’s biological age, while the FAHR-FaceSurvival model predicts survival probability by combining facial features with clinical data.
The models look at distinct parts of the face. Study author Dr. Raymond Mak stated that the age estimation algorithm focuses on the temples, forehead, area above the eyes, and the nasolabial fold connecting the nose to the mouth. In contrast, the mortality risk prediction algorithm examines the bridge of the nose and the area under the eyes.
Echoes of Ancient Prognosis in Modern Algorithms
When evaluating the FAHR-FaceSurvival model among 122 cancer patients flagged as high risk, 52.5% died within 90 days of the assessment. By comparison, only about 2% of the 975 patients in the lowest-risk group died within that timeframe.
Experts Warn That Technology Requires Prospective Clinical Testing
Independent experts emphasize that the technology requires prospective clinical testing.
Miroslaw Bober, a professor of video processing and machine intelligence at the University of Surrey, added that predicting an outcome does not equal explaining it. Bober pointed out that the AI estimates age from appearance rather than measuring underlying biological aging processes directly, and the study did not test whether the AI recognizes anything a physician would miss during a standard examination.
Common Questions About Facial AI Health Assessment
Can cosmetic procedures affect the AI predictions?
The researchers found that cosmetic changes such as surgery, Botox injections, and makeup did not impair the performance of the AI models.
Are these models ready for use in hospitals?
No. Study authors and external reviewers agree that the technology is currently a promising proof of concept and is not ready for clinical use. It must still be tested against doctors’ judgments and across multiple hospital systems.
How can the public participate in ongoing research?
Mass General Brigham is expanding its research through an ongoing online study open to the public.
