Doctors say AI model can predict ‘biological age’ from a selfie — and want to use it to guide cancer treatment
A new artificial intelligence (AI) model can predict a person’s biological age — the state of their body and how they’re aging — from a selfie.
The model, dubbed FaceAge, estimates how old a person looks compared to their chronological age, or the amount of time that’s passed since their birth. FaceAge’s makers say their tool could help doctors decide on the best course of treatment for diseases like cancer. But one outside expert told Live Science that before it is used that way, follow-up data needs to show it actually improves treatment outcomes or quality of life.
When a doctor is treating a cancer patient, “one of the first things they do is they try to assess how well the individual is doing,” Hugo Aerts, director of the AI in Medicine Program at Mass General Brigham, said in a news briefing on May 7. “This is often a very subjective assessment, but it can influence a lot of future decisions” about their treatment, including how aggressive or intense their treatment plan should be, he added. For example, doctors may decide a patient who looks younger and more fit for their age may tolerate an aggressive treatment better and eventually live longer than a patient who looks older and more frail, even if the two have the same chronological age.
FaceAge could make that decision easier by turning doctors’ subjective estimates into a quantitative measure, the study authors wrote in the new study published May 8 in the journal Lancet Digital Health. By quantifying biological age, the model could offer another data point in helping doctors decide which treatment to recommend.
Aerts and his colleagues trained the model on more than 58,000 photos of people ages 60 years and older who were assumed to be of average health for their age at the time the photo was taken. In this training set, the researchers had the model estimate chronological ages and assumed that the people’s biological ages were similar, though the scientists noted that this assumption is not true in every case.
The team then used FaceAge to predict the ages of more than 6,000 people with cancer. Cancer patients looked about five years older, on average, than their chronological ages, the team found. FaceAge’s estimates also correlated with survival after treatment: The older a person looked, regardless of their chronological age, the lower their chances of living longer. By contrast, chronological age was not a good predictor of survival in cancer patients, the team found.
FaceAge isn’t ready for hospitals or physicians’ offices yet. For one, the dataset used to train the model was pulled from IMDb and Wikipedia — which may not represent the general population, and may also not account for factors like plastic surgery, lifestyle differences, or images that have been digitally retouched. Further studies with larger and more representative training sets are needed to understand how those factors impact FaceAge estimations, the authors said.
And the researchers are still improving the algorithm with additional training data and testing its efficacy for other conditions besides cancer. They’re also investigating what factors the model draws on to make its predictions. But once it’s finalized, FaceAge could, for example, help doctors tailor the intensity of cancer treatments like radiation and chemotherapy to specific patients, study co-author Dr. Ray Mak, a radiation oncologist at Mass General Brigham, said during the briefing. A clinical trial for cancer patients, comparing FaceAge to more traditional measures of a patient’s frailty, is starting soon, Mak added.
Ethical guidelines surrounding how FaceAge information can be used, such as whether health insurance or life insurance providers could access FaceAge estimates to make coverage decisions, should be established before rolling out the model, the researchers said. “It is for sure something that needs attention, to assure that these technologies are used only for the benefit of the patient,” Aerts said in the briefing.
Doctors would also need to carefully consider when and how they use FaceAge in clinical settings, said Nicola White, a palliative care researcher at University College London who was not involved in the study. “When you’re dealing with people, it’s very different to dealing with statistics,” White told Live Science. A long-term study assessing whether involving FaceAge in treatment decisions improved patients’ quality of life is needed, she said.
The researchers noted the AI tool wouldn’t be making calls about treatment on its own. “It’s not a replacement for clinician judgement,” Mak said. But FaceAge could become part of a physician’s toolkit for personalizing a treatment plan, “like having another vital sign data point.”
Source link
