© 2025 KRWG
Play Live Radio
Next Up:
0:00
0:00
0:00 0:00
Available On Air Stations

Researchers try AI to help determine how well a patient has aged

File image showing facial recognition technology (Getty Images)
/
File image showing facial recognition technology (Getty Images)

A new artificial intelligence algorithm can scan your face to determine your biological age. That information can be used to determine treatment options for a number of patients, including how aggressively to treat patients with diseases like cancer or other chronic illnesses.

“We tested the medical impact of this technology by applying it across several thousand cancer patients to analyze their face age,” said Dr. Ray Mak, an oncologist at Mass General Brigham and lead author of the study, “And what we found is that patients that looked older had the worst survival outcomes.”

6 questions with Dr. Ray Mak

What does biological age mean?

“So, we typically think of age as a chronological age, the age based on your date of birth. But we also know that people age at differing rates. A biological age would be a measure of how old somebody is based on their physical and physiological state. And what people may not know is that doctors often use biological age to make important decisions.

“I’m an oncologist, and oftentimes we have to make these tough decisions. We’ll use chronological age in what we call risk calculators, decision support tools to decide whether we should offer a tough treatment like chemotherapy or radiation. But then we have to use the subjective assessment, the clinician’s judgment, to understand how healthy is a patient.

“So, a person who has a high biological age looks a lot older than their chronological age, that’s someone where we might have to think about de-intensifying their cancer treatment or providing them with more supportive care to get through the treatment.”

How do you know if it’s accurate?

“We can look across thousands of patients that we presume are healthy and look at the correlation between the chronological age and biological age. There is some variation, but in a healthy population, that variation that we’re seeing, the face age is accurate within plus or minus four years.

“So, it’s not a tool that’s gonna allow you to say like, ‘Am I 35 or 35 and a half?’ But it’s a tool that allows you to say, ‘This is a 35-year-old that looks like a 50-year-old. There might be some problems here.’”

Should doctors be relying solely on technology like this or is it just another tool in the toolbox?

“I think of this as another tool in the toolbox, so that just like vital signs, blood tests, medical imaging, we now have another quantitative data point for doctors that use in their decision making.”

What exactly is this technology looking at on the face?

“That’s one of the really hard things about these artificial intelligence technologies. There’s some inherent black-box nature to it. Oftentimes, we don’t know what it’s looking at.”

How could you prevent this technology from being misused?

“Ultimately, we decided to pursue it because we thought it was important for one, that the potential benefits might outweigh those concerns. Many AI technologies are double-edged swords in that manner, but the other important part of why we did this research is to raise awareness that not only is your face reflective of your identity, but also reflective of your health.

“And with that awareness, we hope that regulatory groups, research scientists like us, can create those ethical and regulatory safeguards to prevent that kind of misuse.”

What’s next for technology of this kind?

“I think we have a long road ahead before it’s commonly used in the clinic. Some of the short-term goals we have are to improve the accuracy of the face age predictions.

‘So, we’re now training on 20 million photographs a new algorithm that can predict face age more accurately. And the last piece, which kind of gets into like the science fiction level, next steps is can we develop AI algorithms that not only predict face age, but facial health?

“So, for example, can we predict the likelihood of someone dying or can we predict the presence of a disease like cancer directly from a face photograph?”

This interview has been edited for clarity.

____

Thomas Danielian produced and edited this interview for broadcast with Mark Navin. Grace Griffin adapted it for the web.

This article was originally published on WBUR.org.

Copyright 2025 WBUR

Deborah Becker
Thomas Danielian