AI Predicts Cancer Survival from a Single Photo

A simple photograph of your face could reveal surprising clues about your health and how long you might live, especially if you're battling cancer.

Scientists at Mass General Brigham in Boston built artificial intelligence that predicts cancer survival and biological age using nothing more than a selfie. Published Monday in The Lancet Digital Health, the breakthrough suggests doctors may soon have a powerful new tool to evaluate treatment options for their patients.

"I was pretty surprised to find out that we could pick up signs of mortality risk from just a facial photograph," said Dr. Raymond Mak, a radiation oncologist who led the research. The AI models were trained on more than 40 million facial images spanning diverse ages and backgrounds.

The technology works by analyzing different facial regions. The age estimation model examines temples, forehead areas above the eyes, and the lines connecting the nose to the mouth. The survival prediction model focuses on the bridge of the nose and areas under the eyes.

What's remarkable is that these AI-identified regions match descriptions written by Hippocrates around 400 B.C. In the Book of Prognostics, the ancient Greek physician described "the nose sharp, the eyes sunken, the temples fallen in" as signs of muscle wasting and mortality risk.

"To think that our algorithms might be replicating that wisdom of Hippocrates from millennia ago was mind-blowing to us," Mak said. The AI learned purely from photographs and survival data, with no knowledge of ancient medical texts.

The results proved impressive. Among 122 cancer patients the AI flagged as high-risk, more than half died within 90 days as predicted. Only 2% of the 975 lowest-risk patients died in the same timeframe.

Biological age predictions showed similar accuracy. Twenty-one percent of patients whose biological age exceeded their actual age by more than 20 years died within 90 days. Only 6% of those with biological ages at least 10 years younger than their chronological age died in the same period.

The researchers found that cosmetic changes like Botox, surgery, or makeup didn't fool the AI. The models looked deeper than surface appearances.

Why This Inspires

This technology isn't meant to replace doctors but to empower them with better information. "If we can pick up accelerated aging or higher mortality scores early, we might be able to screen for those or identify patients who may benefit from further testing for serious conditions earlier," Mak explained.

The tool could help physicians make more informed decisions about which treatments might work best for individual patients. It might also guide earlier interventions for people showing signs of accelerated aging, potentially preventing serious conditions before they develop.

While the study focused on cancer patients, researchers believe the technology could eventually help healthy people interested in understanding their longevity and overall health.

Ancient medical wisdom confirmed by cutting-edge AI offers new hope for better, more personalized cancer care.