Researchers develop AI tool to assess cancer patient survival risk from facial images

A research team at Mass General Brigham has developed a machine-learning model capable of predicting cancer survival rates by analyzing patient facial photographs. The study, published in Lancet Digital Health, details the use of FAHR-FaceSurvival, an AI tool trained on 40 million general facial images and over 30,000 clinical records.

In testing, the model demonstrated a correlation between high-risk scores and early mortality. Patients whose images received the highest 10% of risk scores faced a significantly higher mortality rate within three months compared to those with lower scores. The tool focuses primarily on the bridge of the nose and the area below the eyes.

The research builds upon previous work from the same team, which introduced an AI tool to predict biological age through facial biomarkers. By combining survival risk data with biological age assessments, researchers aim to provide clinicians with a more comprehensive overview of a patient's frailty and overall health.

Dr. Raymond Mak, an oncologist at the Mass General Brigham Cancer Institute, stated that the goal of this technology is to assist in personalized treatment decisions rather than replace clinical judgment. The team intends to continue refining this method as a scalable way to evaluate patient readiness for intensive medical interventions like radiation or surgery.