Study links facial features to cancer patient survival

A new study developed an AI model called 'Fahair Face', trained on over 40 million facial images, deriving two tools to estimate age and rank mortality risk. Analyzing the relationship between estimated age and survival among 38,183 cancer patients treated with radiation, it found that appearing 5 years older than one's actual age was associated with a 21% increase in mortality risk. The relationship is statistical, not causal.

A new study by researchers at Mass General Brigham in the US developed an AI model called 'Fahair Face', trained on over 40 million facial images, from which it derived two tools: one to estimate age from appearance, and another to rank patients by mortality risk based on the image. The study, covered by El País and published in The Lancet Digital Health, analyzed the relationship between estimated age and survival among 38,183 cancer patients treated with radiation. Appearing at least 5 years older than one's actual age was associated with a 21% relative increase in mortality risk. Researchers emphasized this is a statistical association, not evidence that facial appearance causes the outcome. The research institution's statement noted the two tools do not extract the same information, and the correlation between their estimates was weak. The facial regions relied upon by each tool also differ. Miroslav Bobr, a professor at the University of Surrey, noted that testing the model on nearly 5,000 patients in the Netherlands is a strength, but stressed that prediction does not explain outcome, and age estimation from the face does not directly measure internal aging processes. The study confirmed the results do not provide clinicians with a ready-made tool to determine a patient's fate, nor do they prove that using the tool improves treatment decisions or outcomes.