“The face reveals stories that conventional metrics often miss.”
Scientists in the United States have developed an artificial intelligence tool that uses a single facial photograph to estimate a patient’s biological age and assess cancer mortality risks. Researchers at the Boston medical research center Mass General Brigham built the technology to help physicians better evaluate patient physical conditions before determining intensive treatments, such as surgery or radiation therapy, according to reporting by the BBC.
FAHR-Face Uses 40 Million Images to Assess Risk
The research team trained their system, designated as FAHR-Face, using more than 40 million facial images. The specific cancer survival risk model was refined using photographs from over 34,000 patients, and the findings were published in the scientific journal The Lancet Digital Health.
The system splits into two mutually supportive tools. One calculates biological age—indicating how old a person appears rather than their chronological age—while the second tool, FAHR-FaceSurvival, evaluates facial features to determine cancer mortality risk levels.
Facial Zones Indicate Biological Age and Survival Risk
Analysis revealed that different facial areas dictate the output of each model. Zones beneath the eyes and around the root of the nose proved particularly significant for survival risk assessments. Meanwhile, the biological age model focused more heavily on the forehead, temples, and the creases surrounding the nose and mouth. The models demonstrated relevance across patients with various cancer types and disease severity stages.
Researchers tested the model using data from American patients as well as an independent group of nearly 5,000 cancer patients in the Netherlands who received radiation therapy, as reported by the BBC.
Four-Year Observation Windows
Additional testing detailed by the British newspaper The Guardian found that patients who appeared at least five years older than their actual age faced a mortality risk roughly 21 percent higher than other patients over a four-year observation period. Conversely, patients who looked at least five years younger experienced a risk 17 percent lower.
Experts Say Technology Needs More Testing for Clinical Use
Independent experts emphasize that the technology remains unready for clinical practice. Researchers still do not fully understand which specific biological or health shifts the artificial intelligence identifies within facial features. Additional testing is required to verify whether these assessments provide doctors with meaningful supplementary data compared to standard health evaluations, and to check model accuracy across diverse ethnic groups and hospital environments.
More on United States
Worth a look