Artificial Intelligence

Study: AI Predicts Cancer Survival from a Single Face Photo

Are facial-age AI biomarkers a cheap, reliable read on who can handle aggressive cancer treatment or just hype being sold as medicine?

3 hours ago·Updated 1 hour ago

Study: AI Predicts Cancer Survival from a Single Face Photo

Story

Above: A woman takes a selfie in Bielefeld, Germany, on June 2, 2021. Image credit: Friso Gentsch/Picture Alliance/Getty Images

The Facts

Researchers at Mass General Brigham published a study in The Lancet Digital Health on Monday about an artificial intelligence tool that predicts a cancer patient's survival chances from a single facial photograph. The scientists claim the findings are a first.
Scientific American
The Business Times
The Straits Times
"We think that using AI, we can capture some of that physiological reserve, risk for mortality, just from a face photo," said Raymond Mak, a radiation oncologist at Mass General Brigham Cancer Institute and one of the study's authors.
BigGo Finance
The Business Times
The Straits Times
The system includes two algorithms: FAHR-FaceAge, which estimates biological age using creases from the nose to the mouth, the forehead and the temples; and FAHR-FaceSurvival, which predicts mortality based on the area below the eyes and the nasal bridge.
The Lancet
Nacionale
The foundation system was trained on more than 40 million facial photographs and then FaceSurvival was fine-tuned on images of 34,363 cancer patients. It was tested on a separate group of 3,820 patients and checked in nearly 4,900 cancer patients in the Netherlands.
The Lancet
Daily Mail
Of the 122 patients given the highest risk scores in the test group of 3,820, 64 died within 90 days, compared with 20 of the 975 patients in the lowest-risk group. The system worked across ages, races, genders and cancer types.
Scientific American
The Business Times
Daily Mail
Among about 38,000 U.S. cancer patients, those whose faces looked at least five years older than their actual age had a 21% higher risk of death over an average four-year follow-up, while those appearing five or more years younger had a 17% lower risk.
The Guardian
Nacionale
The Sun

Sources Split

LEFT

CENTER

RIGHT

ESTAB
CRIT

CENTER

PRO
ESTAB

The Spin

Techno-optimist narrative

Estimating frailty and physical reserve from a quick photo gives oncologists a cheap, consistent read on who can handle aggressive treatment. Training on tens of millions of images and holding steady across lighting, pose and makeup shows this is serious science. Paired with clinical judgment, facial biomarkers can make cancer care more personalized and more accessible.

Independent

Medicalxpress

Techno-skeptic narrative

This is a promising proof of concept, but the hype is running far ahead of the evidence, since a risk ranking from a face photo hasn’t been validated as an individual survival estimate or as a guide to treatment decisions. Doctors already assess frailty at the bedside, and the study never tested whether the score adds anything beyond that judgment. Loose talk about predicting survival is closer to marketing than medicine.

Science Media Centre