What if your face revealed your real age? Not that indicated on your identity papers, but the one that corresponds to the real aging of your cells. Scientists have developed an artificial intelligence capable of estimating the biological age of a person, paving the way for more personalized medicine, according to a study published in the renowned scientific journal The Lancet Digital Health
The real capacity of a patient to endure heavy treatments
Algorithm called deep learning, Faceage trained on tens of thousands of photos. He notably made it possible to determine that cancer patients had on average five years older, from a biological point of view, than healthy people. Scientists are increasingly considering biological age, that is to say the age of cells, differs from chronological age, the existence of an individual.
In their study published in Lancet Digital Health, researchers believe that faceage could help doctors determine the real capacity of a patient to bear heavy treatments without risk. And could opt, if necessary, for less invasive alternatives.
“Our hypothesis is that faceage could be used as a biomarker in the treatment of cancer to quantify the biological age of a patient and help the doctor make these difficult decisions,” explains the co -author of the study Raymond Mak, a Mass General Brigham oncologist, a university hospital affiliated with Harvard.
For example, a 75 -year -old man whose biological age is 65 years and a person aged 60 whose biological age is 70 years. Invasive radiotherapy could actually be appropriate for the first, but risky for the second. The same logic could guide decisions concerning cardiac surgery, hip prostheses or end -of -life care.
An aging at different rhythms
Human beings age at different rhythms depending on their genes, stress, physical exercise and certain habits like smoking or drinking alcohol. If expensive genetic tests can already reveal DNA wear over time, Faceage promises to provide data from a simple selfie.
This tool has trained on 58,851 adult portraits over the age of 60 a priori healthy, selected from public data. It was then tested on 6,196 patients with cancer in the United States and the Netherlands, from photos taken just before radiotherapy.
Result: patients with malignant tumors seem to be on average 4.79 years older than their chronological age. Among cancer patients, the higher this figure, the lower the chances of survival, even after taking into account the official age, sex and type of tumor. And the risk increases sharply for anyone whose biological age exceeds 85 years.
Ethical risks
For faceage, the signs usually associated with old age, such as gray hair or baldness, import less than subtle changes in the muscle tone of the face. The tool thus makes it possible to improve the degree of precision of doctors. Eight of them were invited to examine photos of cancer terminal patients and to guess who would die in the six months.
Their success rate was barely higher than an evaluation at the ladle. But with the faceage data, their predictions have improved significantly. For the anecdote, faceage estimated the biological age of the actor at 43 Paul Ruddin a photo taken when he had 50.
AI tools are sometimes criticized to take an interest in white people. Raymond Mak, the co -author of the study, claims to have noted any significant racial bias in the predictions of faceage. The group also continues to train for a second generation model on 20,000 patients.
But this technological advance is not, as often, without raising concerns: it could seduce insurers or employers who are still trying to assess the risks as well as possible. Co -author of the study and head of AI in the medical programs of Mass General Brigham, Hugo Aerts calls for “ensuring that these technologies are only used in the interest of the patient”.
