AI learned to determine biological age from selfies

Scientists at Mass General Brigham in Boston have developed and conducted the first tests of an artificial intelligence tool called FaceAge, an algorithm designed to determine the biological age of patients from a photograph of their face, The Washington Post reports.

Biological age is an indicator of a person's health status compared to their chronological age. In medicine, it is used to assess the risks of treatment, particularly cancer, and to predict life expectancy.

FaceAge was trained on tens of thousands of images of people over 60, collected from open sources such as Wikipedia and IMDb. This algorithm determines not chronological age, but biological age - that is, overall health, which is crucial when choosing the intensity of treatment, for example, in the case of chemotherapy.

Preliminary testing results showed that the faces of cancer patients looked, on average, five years older than their actual age. And the older the patient looked, the worse the prognosis for survival.

FaceAge was also tested against doctors' assessments. The accuracy of predicting survival over six months using photos and medical data increased to 80% - compared to 74% for doctors' assessments alone and 61% for images alone.

The developers emphasize that FaceAge does not replace a doctor, but can complement the clinical picture. The technology still needs to be tested for bias, adaptation to different types of lighting, makeup, and other changes in appearance.

The algorithm is not yet ready for mass implementation, but could become an additional biomarker in diagnosis and treatment if it receives regulatory approval.