Cancer treatments are very complicated and need to be customized to each patient based on many factors including the type of cancer they have, how bad it is, how old they are, other medical issues, whether the cancer has spread, and much more. Oncologists will try to look at all the different factors when coming up with a treatment plan that is aggressive enough to be effective, but not so aggressive that the treatment is worse than the disease.

A team of scientists at Mass General Brigham (MGB) claim that they have come up with an AI model that is able to take a picture of a patient and determine just how likely they are to survive a cancer treatment. The system works by analyzing what is known as the biological age of the cancer patient.

Biological age is a measurement that looks at the overall health of the body’s cells rather than just the number of years since they were born. Almost everyone has a basic high-level experience with this concept. When you look at someone and hear there age, you almost instinctively think something like, “Oh, they look old for their age,” or, “Wow, they look really young for their age.” This is a simple form of the concept of biological age.

The AI that was developed, however, is able to determine a patient’s biological age using much more than just a surface level stereotype based on things like wrinkles and balding. The system, which they called FaceAge, was used in a new study that was published in the journal Lancet Digital Health. In the study, the researchers found that those who the AI said appeared younger had a better survival rate than those that the app thought looked older.

The AI model was developed and trained by using almost 59,000 pictures of adults over 50 that were taken from public information. This included sources like IMDB, Wikipedia, and more. Then, for the study, they asked the AI to provide an estimate on the age of 6200 cancer patients. They could then use the results to determine that when the AI thought the patient was younger, they had a better survival rate.

This may seem obvious, but it could also be used as an important tool for oncologists. If the study proves correct, for example, an oncologist may recommend a highly aggressive form of radiation treatment to an 80-year-old patient that the AI determined had a biological age of 65 since they would be more likely to tolerate it. On the other hand, a 65-year-old patient with a biological age of 80 might be better treated with something less aggressive.

The system is not yet fully vetted, and there are some real concerns with it. To start with, it was primarily trained using white faces, so it may not be as accurate for those of other ethnicities. Also, there could be a concern that those who have had plastic surgery might throw the responses off. In a New York Times report, Jennifer Miller, the co-director of the program for biomedical ethics at Yale University said:

“I’d be very worried about whether this tool works equally well for all populations, for example women, older adults, racial and ethnic minorities, those with various disabilities, pregnant women and the like.”

Of course, all of those things can be accounted for with the proper training of the AI. Researchers and doctors would have to be careful to ensure that it is properly trained and accurate for all types of people before it could be considered for widespread use.

If you thought that was interesting, you might like to read about a second giant hole has opened up on the sun’s surface. Here’s what it means.