Researchers have developed artificial intelligence (AI) models that can estimate a person’s biological age and predict cancer survival chances from a simple facial photograph. The technology, created by experts at Mass General Brigham in Massachusetts, was trained on a dataset of over 40 million images representing diverse ages and ethnic backgrounds.
Biological age differs from chronological age by reflecting the physical condition and functional health of an individual rather than the actual number of years lived. The research produced two AI tools: FAHR-FaceAge, which estimates biological age, and FAHR-FaceSurvival, which predicts survival risk for cancer patients based on facial features. While both tools derive from a shared foundation model, they capture distinct but complementary information—the first assesses biological ageing signs primarily from the forehead and temple area, and the second focuses on the region below the eyes and around the nasal bridge.
The study, published in the journal Lancet Digital Health, involved analysis of 34,000 photos of cancer patients. Those classified in the highest risk quartile by the AI were three times more likely to die during the study period than patients in the lowest risk group. Furthermore, among cancer patients treated with radiotherapy in the US, individuals who looked five years or more older than their actual age had a 21% higher mortality risk during an average four-year follow-up. Conversely, patients who appeared younger had a 17% lower risk.
Lifestyle factors such as heavy alcohol consumption, drug use, and smoking were strongly linked to facial ageing in the AI’s assessment. Smoking, in particular, showed the most pronounced association with appearing biologically older, with current smokers looking older than former smokers, who in turn looked older than those who never smoked. The FAHR-FaceAge model demonstrated stability and consistency across common variations in photos, including differences in lighting, facial expressions, makeup, and even some plastic surgery procedures like rhinoplasty.
Researchers highlighted the potential clinical applications of the technology. As populations age, physicians need better tools to estimate patients’ physical reserve and frailty before recommending intensive cancer treatments such as surgery, radiation, or systemic therapy. The AI models could provide quantitative, scalable, and consistent assessments to support personalized treatment decisions. The models might also serve as non-invasive tools to track lifestyle changes or the effects of anti-ageing interventions, offering immediate feedback through facial analysis.
Despite its promise, developers caution that further research is necessary before the AI tool can be widely integrated into clinical practice. The team is currently expanding their research through the one FaceAge Research Study, which aims to enroll 10,000 participants from diverse racial, ethnic, age, and medical backgrounds. Participants can upload photos to receive an estimated biological age as part of the ongoing study. The project represents a significant step in using AI facial recognition to support healthcare decisions, although experts emphasize that these tools are intended to augment—not replace—physician judgment.