Mass General Brigham AI tool predicts cancer survival from facial photographs

Researchers at Mass General Brigham developed an AI model trained on over 40 million facial images that predicts cancer survival from a single photograph. Published in The Lancet Digital Health, the FAHR-Face system powers two tools: FAHR-FaceAge, which estimates biological age, and FAHR-FaceSurvival, which analyzes features around the eyes and nasal bridge to assess mortality risk. In validation trials, high-risk patients were over three times as likely to die within 90 days. Clinicians aim to use the technology to assess frailty and personalize treatment, though independent experts stress further validation is required.

Overview of the FAHR-Face tools

• ▪The FAHR-FaceAge tool estimates a patient's biological age from a photo, while FAHR-FaceSurvival predicts cancer survival outcomes directly from a facial photograph.
• ▪Mass General Brigham researchers developed FAHR-Face, an artificial intelligence foundation model that powers two medical prediction tools called FAHR-FaceAge and FAHR-FaceSurvival.

Validation testing of FAHR-FaceSurvival

• ▪In a validation trial of 3,820 cancer patients, 64 of the 122 patients assigned the highest FAHR-FaceSurvival risk scores died within 90 days, compared to 20 of 975 in the lowest-risk group.
• ▪Cancer patients in the highest-risk quartile of FAHR-FaceSurvival scores were more than three times as likely to die as those in the lowest-risk quartile after accounting for age, sex, and cancer type.
• ▪Researchers validated the FAHR-FaceSurvival prediction model in an external cohort of nearly 4,900 cancer patients in the Netherlands.

Factors affecting FAHR-FaceAge estimates

• ▪The FAHR-FaceAge tool demonstrated biological age estimation correlation scores across racial groups of 0.86 for White, 0.81 for Black, 0.86 for Asian, and 0.91 for Hispanic individuals.
• ▪Heavy alcohol use, drug use, and smoking were associated with looking biologically older in FAHR-FaceAge facial photograph analyses, with current smokers appearing older than former smokers.
• ▪FAHR-FaceAge estimates remained stable across variations in photo pose, lighting, expression, makeup, and most plastic surgery procedures, showing only a slight shift after rhinoplasty.

Mechanisms comparing the prediction tools

• ▪FAHR-FaceAge and FAHR-FaceSurvival yield weakly correlated predictions, and combining both measures improved survival-model fit in external validation testing.
• ▪FAHR-FaceSurvival directly learns cancer mortality risk from facial photographs without inferring survival through an intermediate biological age calculation.
• ▪FAHR-FaceSurvival evaluates the facial region below the eyes and around the nasal bridge to predict mortality risk, whereas FAHR-FaceAge evaluates nasolabial folds, the forehead, and temples.

Expert perspectives on clinical potential

• ▪Mass General Brigham radiation oncologist Dr. Raymond Mak stated that the AI facial tools aim to quantify patient physical reserve and frailty to support personalized treatment decisions.
• ▪University of Surrey machine learning expert Prof. Miroslaw Bober cautioned that the technology remains a proof of concept that requires further testing across hospitals before clinical deployment.
• ▪Colon and rectal surgery expert Dr. Grace Lee stated that the FaceSurvival tool provides a standardized method to quantify clinicians' intuitive visual assessments of patient frailty.

Model training datasets

• ▪Mass General Brigham researchers trained the foundational FAHR-Face model on more than 40 million facial photographs across diverse ethnicities and ages.
• ▪Researchers trained and validated the FAHR-Face survival prediction tools using photographs of over 30,000 cancer patients with known medical outcomes.

Patents and ongoing research

• ▪Mass General Brigham researchers filed a patent application for the FaceSurvival artificial intelligence tool.
• ▪Mass General Brigham researchers expanded their work through an online, institutional review board-approved public study where participants upload facial photographs to receive biological age readouts.