How a photo of a person’s face could help predict whether they’ll survive cancer treatment

Key Points:

• Researchers at Mass General Brigham have developed AI models that can predict cancer patients' biological age and survival chances using just a single facial photograph, potentially aiding doctors in treatment decisions.
• The AI assesses specific facial regions linked to biological aging and mortality risk, echoing observations made by Hippocrates over 2,000 years ago, despite the algorithms being trained solely on images and patient outcomes.
• In studies, the AI models showed significant predictive accuracy: over half of high-risk patients identified by the model died within 90 days, while low-risk patients had much lower mortality rates; biological age estimates also correlated with survival outcomes.
• Cosmetic alterations like surgery or makeup did not affect the AI's performance, but researchers stress the technology is an emerging tool meant to complement, not replace, clinical evaluations.
• Experts caution that while the findings are promising, the AI’s clinical utility requires prospective testing across diverse populations and comparison with doctors’ assessments before it can be widely adopted.