AI Tool Predicts Cancer Survival From a Selfie

Mass General Brigham researchers built a foundation AI model called FAHR‑Face. According to The Guardian, the model was trained on more than 40 million facial photographs. The study was published in Lancet Digital Health. The authors include Hugo Aerts and Raymond Mak.

According to MedicalXpress, FAHR‑Face learns a general representation of the face. The model can be adapted to different health outcomes. The study used the model to create two tools. One tool estimates biological age; the other predicts cancer survival.

According to MedicalXpress, FAHR‑FaceAge focuses on the nasolabial folds, the forehead above the eyes, and the temples. According to MedicalXpress, FAHR‑FaceSurvival evaluates the area below the eyes and around the nasal bridge. According to MedicalXpress, the two tools rely on different facial regions. According to MedicalXpress, their predictions are only weakly correlated.

According to MedicalXpress, the study found that patients whose FAHR‑FaceAge was five or more years older than their chronological age had a 21 % higher risk of death during a four‑year follow‑up. According to MedicalXpress, patients who looked five or more years younger had a 17 % lower risk. According to MedicalXpress, the deviation between FAHR‑FaceAge and chronological age was independently linked to survival.

According to MedicalXpress, researchers also examined lifestyle factors. According to MedicalXpress, heavy alcohol use, drug use, and smoking were associated with looking biologically older. According to MedicalXpress, current smokers appeared older than former smokers, and former smokers older than never smokers. According to MedicalXpress, smoking had the strongest pattern.

According to Raymond Mak, physicians need better ways to estimate physical reserve before recommending intensive treatments. According to MedicalXpress, the authors say the AI tools can support personalized treatment decisions but do not replace physician judgment. According to MedicalXpress, they suggest the approach is scalable and quantitative.