Mass General Brigham team study: AI looks at faces to predict survival risk of cancer patients

Can a single facial photo reveal how much life a person has left?

Of the 122 people with the highest scores, 64 died within 90 days.

Of the 975 people with the lowest scores, over the same 90 days, only 20 died.

There was nothing special about these photos; most were routine pictures taken by the hospital before patients underwent radiotherapy.

Ray Mak is a radiation oncologist at Mass General Brigham in the United States and one of the leaders of this study. He says this is how medical school trains doctors: walk into the ward, take one look at the patient, and get a sense of their overall condition.

That glance matters.

Ray Mak once treated an 86-year-old lung cancer patient who looked considerably younger than his actual age, which was one of the reasons Ray Mak recommended more aggressive treatment. By last year, the man was already 90 and still doing well.

Later, researchers used AI to analyze his photo from back then, and the estimated age came out a full 10 years younger than his actual age.

The reverse is also true. Some people look frailer than their peers, and their bodies may not be able to withstand overly aggressive treatment.

The problem is that this glance relies entirely on experience, and human eyesight has never been very reliable.

In the team's first-generation study published last year, 10 doctors and researchers looked at photos of 100 patients with advanced cancer and judged their short-term survival. Looking only at the photos, their accuracy was about 60%, only slightly better than random guessing; after receiving the AI's analysis results, it rose to about 80%.

This time, they wanted the AI to see more accurately. On October 19, the new study was published in the medical journal The Lancet Digital Health.

First, look at 40 million faces

With the foundation in place, the researchers began teaching it to look at health.

They first taught it to estimate "biological age." Biological age refers to the age of the body, not the age on an ID card. Two people who are both 60 may have very different biological ages: one can climb five flights of stairs without panting, while the other gets tired after a few steps.

On the same foundation, they also trained a bolder version: estimating mortality risk directly from faces. To do this, the researchers fed it photos of more than 34,000 cancer patients, along with their subsequent survival outcomes, to learn from.

The 3,820 people mentioned at the beginning were deliberately held out as "examinees" that the AI was not allowed to study in advance.

As a result, neither version disappointed.

The biological age version was no slouch either. The researchers used it to look at all of the more than 38,000 radiotherapy patients and found that people judged by it as "looking more than 5 years older than their actual age" still had a 21% higher risk of death, even after accounting for factors such as age, sex, and cancer type.

Clues humans cannot read

The AI's "attention maps" when predicting survival risk (left) and age (right); brighter colors mean the AI is paying more attention to that area | Mass General Brigham

Is it eye bags? Wrinkles? Pigmentation? Or skin texture? At present, it cannot be determined.

Schematic of the AI process for assessing biological age from facial photos | Mass General Brigham

The weight of a selfie

The team has already begun trying it in hospitals. They take photos of patients over 65 with early-stage lung cancer who are about to receive radiotherapy, to see whether the AI can identify those who are relatively frail and need extra nutritional support or rehabilitation training.

It could also serve as a kind of early warning: if a person's face starts aging faster than those of their peers, it might mean something is happening in the body.

The general public can already use it too. In March this year, the team quietly launched a research website where anyone can upload a few selfies and get the AI's estimated biological age within seconds. So far, more than 10,000 people have signed up. The researchers stress that the results are for research only, not diagnosis.

These selfies will in turn help the AI continue learning. This is exactly what it needs most right now: the AI predicting mortality risk was trained on photos of patients in the Boston area, most of whom are white. Whether it is still accurate with a different group of people is hard to say for now. The team hopes that photos uploaded by the public will let the AI see more different faces.

What is harder to deal with than the technology are the issues beyond it.

Matthew Liao, director of the Center for Bioethics at New York University in the United States, believes that such tools may help doctors identify patients who need extra care; but AI can also make mistakes, and once a person who needs treatment is misjudged as "fine," that person may not receive the care they deserve.

He also worries about another scenario: insurance companies using it to decide whether to insure someone, or employers using it to pick the one who looks "healthier" from a group of job applicants.

A facial photo takes only a second to take, and almost everyone has one on their phone. This is exactly the benefit of this technology, and also what makes people uneasy about it.

Matthew Liao says that while advancing this technology, we really need to think clearly about its impact on society.

References

[2] Haugg F, et al. Foundation artificial intelligence models for health recognition by use of face photographs (FAHR-Face): a model development and validation study. The Lancet Digital Health, 2026. DOI: 10.1016/j.landig.2026.101056

[3] Mass General Brigham. AI facial health recognition tools help physicians assess age and survival risk to guide personalized treatment. 2026-10-19. https://news.massgeneralbrigham.org/en/ai-facial-recognition-tools-assess-age-cancer-survival-risk

[4] Bontempi D, et al. FaceAge, a deep learning system to estimate biological age from face photographs to improve prognostication: a model development and validation study. The Lancet Digital Health, 2025. DOI: 10.1016/j.landig.2025.03.002

[5] Rahal S. You can now upload a selfie to a Mass General Brigham AI tool and it’ll estimate your biological age. The Boston Globe, 2026-10-20. https://www.bostonglobe.com/2026/10/20/metro/ai-biological-age-face-selfie-free-tool/

This article is from the WeChat public account "Guokr" (ID: Guokr42), author: Steed, published by 36Kr with authorization.