Researchers at Mass General Brigham in Massachusetts created an artificial intelligence system capable of predicting cancer patient survival using a single facial photograph. The study, published in the journal The Lancet Digital Health, introduces a tool named FAHR-Face.
FAHR-Face operates by analysing facial features to estimate mortality risk and biological age. More than 40 million facial photos were used to train the underlying model.
Scientists adapted the technology into two specific versions: FAHR-FaceAge for estimating biological age and FAHR-FaceSurvival for predicting cancer outcomes. The survival model specifically focuses on areas around the nose and beneath the eyes, while the age-estimation model examines the temples and mouth.
Training for the survival model involved images of 34,363 patients, which were then tested on a group of 3,820 individuals. Independent validation was conducted using data from 4,892 cancer patients in the Netherlands.
Patients identified in the highest-risk category faced more than triple the mortality risk of those in the lowest-risk group. Individuals whose faces appeared at least 5 years older than their actual age showed a 21% higher mortality risk, whereas those appearing 5 years younger had a 17% lower risk.
Combining both models increased the accuracy of survival predictions. Researchers stated the technology could help doctors assess patient health and guide treatment decisions.
