AI predicts cancer survival from facial photos

Researchers at Mass General Brigham in Boston have developed artificial intelligence models that can predict a cancer patient's biological age and survival chances from a single facial photograph, according to a study published Monday in The Lancet Digital Health. The AI models, trained on over 40 million facial images, assess specific facial regions to gauge health and mortality risk. The study found that among 122 cancer patients flagged as high-risk by the AI model, 52.5 percent died within 90 days, compared to about two percent of the 975 lowest-risk patients. The researchers noted that cosmetic changes like makeup or Botox did not affect the models' performance, and the facial regions analyzed echo descriptions from Hippocrates' ancient medical texts. Experts called the findings a promising proof of concept but cautioned that the technology requires prospective clinical testing and validation against doctors' own assessments before it can be used in practice. The authors emphasize the AI is not intended to replace physicians but could serve as an additional screening tool.