AI tool predicts cancer survival and biological age from a single facial photo

Mass General Brigham researchers trained the model on 42 million face images and have opened an expanded study to the public

By LineZotpaper AI staff

Published26 October 2026

Read Time2 min

Follow this storyHear when it moves

Illustration: AI tool predicts cancer survival and biological age from a single facial photo

Researchers at Mass General Brigham Cancer Institute say they have built an AI tool that can estimate a patient's biological age and their chances of surviving intensive cancer treatment from a single photograph of their face, in what the team describes as a first for AI in medicine.

In research published in The Lancet Digital Health, the team describes FAHR-Face, a foundation model trained on more than 42 million facial photos using a self-supervised method called masked autoencoders to build a holistic understanding of facial structure. From that model they developed two tools: FAHR-FaceAge, which estimates biological age, and FAHR-FaceSurvival, which directly predicts survival outcomes for cancer patients.

The researchers say FAHR-FaceAge estimated chronological age with an average error of 5.1 years, compared with 6.92 years for another existing model, and was not thrown off by makeup, glasses, facelifts or changes in lighting, though it made small errors on images of people who had undergone nose jobs. Patients who appeared five or more years older than their actual age had a 21% higher risk of mortality, while looking five years younger was associated with a 17% higher survival rate. FAHR-FaceSurvival, trained on photos of more than 34,000 radiotherapy patients plus two smaller groups of nearly 4,000 and 5,000 patients, identified high-risk patients as having more than triple the mortality rate of low-risk patients; 52% of the highest-risk group died within 90 days, compared with 2.1% in the lowest-risk group.

The two models draw on different facial regions: FaceSurvival evaluates the area below the eyes and around the nasal bridge, while FaceAge focuses on the nasolabial folds, the forehead area just above the orbits and the temples. The team says consumer-grade facial photos can serve as a reliable, non-invasive biomarker to help clinicians gauge patients' overall health and treatment risk. A larger study is open to the public, and participants can submit a photo and receive their own FaceAge readout.

RSS

The Daily Edition

The front page in your inbox each weekday morning: the lead, the picks, and why each one earned its place.

Analysis

Why This Matters

The tool could give clinicians a quick, non-invasive way to estimate physical reserve and frailty before recommending intensive treatments such as radiation or surgery.
It may reduce the need for rigorous testing in older patients who are already unwell, according to the researchers.
The public study could broaden the data behind the model and test how well it works beyond the original patient groups.

Background

Chronological age is an imperfect guide to how well someone will tolerate cancer treatment. Patients of the same age can have very different levels of physical fitness and resilience, and doctors have traditionally relied on clinical judgment when making these calls. This study is part of a broader push to use AI foundation models to develop health biomarkers from ordinary images rather than training each model from scratch.

Key Perspectives

Raymond Mak: says that as the population ages, physicians need better ways to estimate physical reserve and frailty before recommending intensive treatments. Hugo Aerts: says this is, to the team's knowledge, the first demonstration that an AI model can directly predict cancer survival from a simple photograph, and notes that the foundation-model approach makes it scalable to develop multiple facial health biomarkers.

What to Watch

Whether data gathered through the public photo study improve or change the accuracy of the FaceAge and FaceSurvival models.
Whether independent validation or clinical adoption follows the publication in The Lancet Digital Health.
How the models perform across more diverse patient populations and a wider range of photo conditions.

Sources

Single photo predicts your chances of surviving cancer treatment · New Atlas

Checked

12 claims in this story checked against 1 source: 10 supported.

Show each claim

SupportedThe AI tool predicts biological age and cancer-treatment survival from a single facial photo.New Atlas: Technology-Innovation-Outdoor News: “Researchers at Mass General Brigham Cancer Institute have developed an AI tool that determines your biological age and your chances of surviving challenging cancer treatment – all from a single photograph of your face.”
Partly supportedThe model was trained on 42 million facial images.New Atlas: Technology-Innovation-Outdoor News: “Using a self-supervised learning method called masked autoencoders, it was trained on more than 42 million photos of faces”
SupportedAn expanded study is open to the public, and participants can submit a photo and receive a FaceAge readout.New Atlas: Technology-Innovation-Outdoor News: “you can participate in an expanded study the team has opened to the public via this page, and receive your own FaceAge readout.”
Partly supportedThe team describes the tool as a first for AI in medicine.New Atlas: Technology-Innovation-Outdoor News: ““To our knowledge, this is the first demonstration that an AI model can directly predict cancer survival from a simple photograph,” said Hugo Aerts”
SupportedResearchers at Mass General Brigham Cancer Institute built a tool that estimates biological age and chances of surviving intensive cancer treatment from one photo.New Atlas: Technology-Innovation-Outdoor News: “Researchers at Mass General Brigham Cancer Institute have developed an AI tool that determines your biological age and your chances of surviving challenging cancer treatment – all from a single photograph of your face.”
SupportedThe research was published in The Lancet Digital Health.New Atlas: Technology-Innovation-Outdoor News: “This research was published in The Lancet Digital Health”
SupportedFAHR-FaceAge estimated chronological age with an average error of 5.1 years, compared with 6.92 years for another model.New Atlas: Technology-Innovation-Outdoor News: “The FaceAge model estimated people's chronological age more accurately than other existing models out there, with a low average error of 5.1 years (compared to another model's average error of 6.92 years).”
SupportedPatients who appeared five or more years older than their actual age had a 21% higher risk of mortality.New Atlas: Technology-Innovation-Outdoor News: “Patients who appeared five or more years older than their actual age had a 21% higher risk of mortality compared to those who looked their actual age.”
SupportedLooking five years younger than one's actual age was associated with a 17% higher survival rate.New Atlas: Technology-Innovation-Outdoor News: “Conversely, looking five years younger than one's real age was associated with 17% higher survival rates.”
SupportedFaceSurvival was trained on photos of more than 34,000 radiotherapy patients and two groups of nearly 4,000 and 5,000 patients.New Atlas: Technology-Innovation-Outdoor News: “FaceSurvival, was further trained on facial photographs of more than 34,000 cancer patients undergoing radiotherapy, as well as two separate groups of nearly 4,000 and 5,000 cancer patients, respectively.”
SupportedPatients in the highest-risk category had more than triple the mortality rate of those in the lowest-risk category.New Atlas: Technology-Innovation-Outdoor News: “those it placed in the highest-risk category had more than triple the mortality rate, compared to those in the lowest-risk category.”
SupportedIn the highest-risk group, 52% died within 90 days, compared with 2.1% in the lowest-risk group.New Atlas: Technology-Innovation-Outdoor News: “in the highest-risk score group, 52% of patients died within 90 days, compared to only 2.1% in the lowest-risk group.”