Researchers at Mass General Brigham in Boston have developed an artificial intelligence model that can assess cancer patients’ chances of survival using only a photograph of their face. The study, published in The Lancet Digital Health, indicates significant implications for personalized treatment.
Dr. Raymond Mak, a radiation oncologist and the study’s lead author, said the AI models evaluate facial features to determine both biological age and mortality risk. The technology was trained on over 40 million images to enhance its accuracy across diverse demographic backgrounds.
The researchers created two algorithms: FAHR-FaceAge for estimating biological age and FAHR-FaceSurvival for predicting survival outcomes. Each algorithm analyzes different facial regions to derive its assessments. For example, FAHR-FaceAge focuses on features like the area around the temples and nasolabial folds, while FAHR-FaceSurvival concentrates on the bridge of the nose and areas under the eyes.
The study found that cancer patients who appeared biologically older than five years compared to their chronological age faced a higher mortality risk, with more than 21% dying within 90 days of treatment. In contrast, only 2% of those who looked younger than their chronological age faced the same fate.
This tool is part of a larger vision to enhance health assessments, particularly for older patients at risk for complications. Ongoing clinical trials aim to validate its effectiveness by enrolling participants across various treatment paradigms.
Public Participation and Future Testing
A web-based platform, FaceAge, has been launched, allowing individuals to upload selfies to receive an estimate of their biological age. Since its introduction, over 10,000 people have participated, aiding researchers in correlating lifestyle choices with facial aging.
Experts recognize the study’s potential but stress that further validation is crucial. Dr. Wilbur Lam, a research chair at Emory University, remarked on the compelling nature of the findings but highlighted the need for prospective evaluation for clinical utility.
Technical Insights and Limitations
Despite being a notable advancement, the AI technology is not yet intended to replace comprehensive medical assessments. The research team continues to investigate how factors like lighting and facial expression can influence predictions. They noted that while most cosmetic procedures have little impact, certain changes could slightly adjust biological age estimates.
The AI’s ability to distinguish between various facial features without explicit training on such metrics has been described as both exciting and eerie, echoing assessment methods attributed to Hippocrates. However, caution remains regarding its readiness for widespread use.
