Researchers from Mass General Brigham have introduced FAHR-Face, a set of neural networks capable of estimating biological age and survival risks for cancer patients based on a single photograph. The study, published in The Lancet Digital Health, details how the model was trained on over 40 million images using a visual transformer architecture.
The system consists of two specialized versions. FAHR-FaceAge determines biological age, showing an average error of 5.1 years in tests. FAHR-FaceSurvival evaluates survival probabilities for cancer patients, processing a single image in less than five seconds on a standard processor.
Clinical tests involving 38 183 patients showed that individuals appearing five years older than their chronological age faced a 21% higher mortality risk. Conversely, those appearing younger had a 17% lower risk. The models analyze specific facial zones, such as the area under the eyes and the bridge of the nose.
The developers released the code under an Apache 2.0 license, though weights for FAHR-FaceSurvival remain proprietary. Mass General Brigham has filed a patent application for these health-assessment algorithms. Researchers are currently conducting an online study to further validate the technology on a broader population.
