Researchers at Mass General Brigham have built an artificial intelligence system that estimates biological age and directly predicts cancer survival outcomes from a single facial photograph. Detailed in the Lancet Digital Health, the foundation model was trained on more than 40 million facial images across diverse ethnicities and ages, marking a significant advancement in facial health recognition technology.
The system yielded two complementary tools: FAHR-FaceAge, which estimates biological age and predicts cancer survival, and FAHR-FaceSurvival, which predicts survival risk directly from facial photos. The algorithms evaluate distinct regions of the face; for instance, FAHR-FaceAge focuses on facial creases and forehead areas, while FAHR-FaceSurvival evaluates the regions below the eyes and around the nasal bridge.
In a study of US cancer patients treated with radiotherapy, individuals evaluated by FAHR-FaceAge as looking five or more years older than their chronological age had a 21% higher risk of dying during a four-year follow-up period. Researchers expect these scalable facial health biomarkers to assist physicians in predicting surgical complications and determining personalized patient care strategies.
Why it matters
Demonstrates foundation models trained on non-traditional datasets can yield scalable biological markers for predictive healthcare applications.
Offers clinicians non-invasive risk-stratification tools to guide treatment decisions and predict post-surgery or post-radiotherapy outcomes.
Source: theguardian.com



