AI Tool Can Predict Cancer Survival From a Face Photo

Scientists have built an AI tool that can look at a photo of a cancer patient's face and predict survival chances. The tool was trained on more than 40 million face images and photos of over 30,000 cancer patients. It gave each patient a risk score, and those with the highest scores were more likely to die within 90 days. The study was published on October 19 in a medical journal called The Lancet Digital Health.

The model could also estimate a person's age with a margin of error of about five years. It worked across different ages, races, genders, and types of cancer. This showed that the tool could be useful for a wide range of patients.

Raymond Mak is a cancer doctor at Mass General Brigham Cancer Institute and helped build the tool. He said doctors are trained to judge health just by looking at a patient, but they are not very good at estimating how long someone has to live. Mak said he was stunned by how well the AI predicted survival.

When the team looked more closely, they found the AI focused on the bridge of the nose and the area just below the eyes. Mak discovered that the ancient Greek doctor Hippocrates had described those same facial features as early signs of approaching death. This link between ancient medicine and modern AI amazed the research team.

Not everyone agrees the tool is ready for real-world use. Professor Miroslaw Bober of the University of Surrey said predicting an outcome is not the same as explaining it. He said the tool still needs to be tested against doctors' own judgment and across different hospitals and patient groups.

Mak agreed that more testing is needed, but he believes the tool shows real promise. He said it is meant to give doctors one extra piece of information, not to take over their decisions or limit what care a patient gets. AI tools like this one are part of a growing effort to use technology to help doctors detect and treat cancer more effectively.