Artificial intelligence learned to identify signs of near death – using selfies

Researchers at Massachusetts General Brigham have developed AI models that predict biological age and survival chances of cancer patients from a single facial photo. 52.5% of patients identified as high-risk died within 90 days. Experts call it a promising proof of concept.

Researchers at Massachusetts General Brigham Hospital have developed artificial intelligence models capable of predicting the biological age and survival chances of cancer patients using a single facial photo. The system, trained on over 40 million facial images, identifies specific areas of the face associated with mortality risk. The study found that among 122 patients identified as high-risk, 52.5% died within 90 days, compared to about 2% in the low-risk group. Researchers discovered that the facial areas the AI focuses on overlap with descriptions by Hippocrates from 400 years ago. External experts call the study "a promising proof of concept" but emphasize that further clinical trials are needed.