A study published in The Lancet Digital Health explores how AI could help assess biological aging and mortality risks in cancer patients, while experts stress that clinical validation is still needed.
Researchers at the Mass General Brigham Institute in Boston have developed an AI model capable of analyzing a single facial image to estimate biological age and mortality risks among cancer patients, potentially paving the way for new tools to assess health status and guide treatment decisions.
The study, published in The Lancet Digital Health, is based on models trained on more than 40 million facial images, with the aim of identifying markers associated with biological aging and survival probabilities.
Facial analysis technology remains a research tool under evaluation (Pexels)
Biological age differs from chronological age, reflecting indicators related to the body's condition and aging rather than the number of years a person has lived.
The researchers used two separate models: one to estimate biological age and another to predict mortality risks in cancer patients based on facial features and additional clinical information.
The results showed that 52.5% of patients classified by the survival prediction model as belonging to the highest-risk category died within 90 days of assessment, compared with about 2% in the lowest-risk group.
Despite these remarkable results, researchers and independent experts emphasized that the technology is not intended to replace doctors or medical tests. Its ability to improve treatment decisions still requires future clinical evaluation.
The model's accuracy across different age groups, skin colors, and facial features remains under study. Experts warn that a statistical association between facial features and mortality risks alone does not directly prove that an image reveals the mechanisms underlying biological aging.
Thus, the technology represents a promising research tool, not a ready-to-use system for determining patient outcomes or dictating treatment.
Thank you for subscribing to our newsletter
