AI predicts cancer survival from a selfie

FaceAge, a new AI tool developed by Mass General Brigham, can predict cancer survival more accurately than clinicians – using only a facial photo.

Brigham researchers found that patients with cancer appeared about five years older than their chronological age and were associated with worse overall survival outcomes.

FaceAge estimates biological age and cancer survival outcomes based on a photo, and is better at predicting short-term life expectancies of patients receiving palliative radiotherapy than clinicians, according to researchers.

Face pictures can provide “clinically meaningful” information, according to co-senior and corresponding author Hugo Aerts, PhD, director of the Artificial Intelligence in Medicine (AIM) at Brigham.

“This work demonstrates that a photo like a simple selfie contains important information that could help to inform clinical decision-making and care plans for patients and clinicians,” Aerts said.

“How old someone looks compared to their chronological age really matters – individuals with FaceAges that are younger than their chronological ages do significantly better after cancer therapy.”

The results of the study were published in The Lancet Digital Health.

Why does AI see what doctors can’t?

While a patient’s appearance may give physicians clues about their overall health and help determine the best course of treatment, doctors, like anyone else, may have biases that could influence them.

A machine does not and, as this study indicates, can provide a more objective assessment. In tests, clinicians were asked to predict short-term survival based on patient photos. Their predictions were only “slightly better than a coin flip,” even after they were given clinical content, according to researchers.

The AI, by contrast, provided more accurate assessments. FaceAge was trained on nearly 60,000 photos of healthy individuals and tested on over 6,000 images of cancer patients from two medical centers. Results showed a correlation between looking older and lower survival rates, particularly among those who appeared 85 or older.

While difficult to pin down exactly, the survival estimate is important in terms of treatment. Further research is needed, but this “opens the door to a whole new realm of biomarker discovery from photographs,” said co-senior author Ray Mak, MD, a faculty member at AIM.

Potential applications also go beyond cancer care or predicting age. “As we increasingly think of different chronic diseases as diseases of aging, it becomes even more important to be able to accurately predict an individual’s aging trajectory,” Mak said.

He added: “I hope we can ultimately use this technology as an early detection system in a variety of applications, within a strong regulatory and ethical framework, to help save lives.”