Researchers have developed a technology that analyzes facial features to estimate cancer patients’ chances of survival. However, facial images cannot predict the future with absolute accuracy.
اضافة اعلان
An ordinary facial photograph may contain health indicators that are difficult to detect with the naked eye. Researchers at Mass General Brigham have developed an artificial intelligence system that analyzes facial images to
estimate cancer patients’ biological age, as well as their likelihood of survival.
The findings, published in The Lancet Digital Health, suggest that facial analysis could eventually help doctors make more informed treatment decisions by providing additional indicators of patients’ health.
How does AI predict cancer survival prospects from facial features?
The researchers developed a foundational AI model trained on more than 40 million facial images of people of different ages and ethnic backgrounds.
They then adapted the model to create two tools, each serving a different purpose.
The first tool estimates a person’s biological age based on their facial appearance the age they appear to be compared with their actual age and uses this estimate to predict their chances of survival after a cancer diagnosis.
The second tool directly estimates the risk of death by analyzing a facial image, without relying on apparent age as an essential step.
Each tool focuses on different areas of the face. The first analyzes wrinkles around the nose and mouth, as well as the forehead and temples, while the second examines the area beneath the eyes and around the bridge of the nose.
The AI models showed promising results across different types of cancer and diverse population groups.
The study found that patients classified by the facial-analysis tool as being in the highest-risk quartile were more than three times as likely to die as those classified in the lowest-risk quartile.
This association remained statistically significant even after the results were adjusted for several clinical factors related to patients’ health.
The biological-age estimation tool also identified another pattern. Cancer patients who appeared at least five years older than their actual age faced a 21% higher risk of death during the four-year follow-up period.
By contrast, the risk of death was 17% lower among patients who appeared at least five years younger than their actual age.
What else did the study reveal?
The researchers found an association between excessive alcohol consumption, drug use and smoking, and signs of accelerated biological aging in the facial images analyzed by the system.
Another notable finding was that the tool maintained its accuracy despite differences in lighting conditions and the use of cosmetics, and even after most cosmetic procedures.
Only a slight change in the results was observed among people who had undergone rhinoplasty.
Nevertheless, the researchers stressed that the findings reflect statistical risks observed across groups of patients. They do not mean the system can accurately determine how long an individual patient will live.
Dr. Raymond Mak, a radiation oncologist and one of the study’s contributors, told the British newspaper The Guardian that the technology is intended to help doctors assess whether patients are physically able to tolerate intensive treatments, such as radiotherapy or surgery, before recommending them.
The researchers emphasized that the AI tool was designed to support medical assessments and help doctors make decisions, not to replace their expertise or clinical judgment.
Despite the encouraging findings, the technology requires further research and testing before it can be adopted widely in medical practice.
