An artificial intelligence (AI) model can estimate biological age and predict survival chances among cancer patients from a simple photograph of their face, according to a study published in The Lancet Digital Health.Researchers at Mass General Brigham, a leading US biomedical research organisation and principal teaching affiliate of Harvard Medical School, developed the model, called FAHR-Face (Foundation Artificial Intelligence for Health Recognition), using more than 40 million publicly available facial images. They then developed two complementary tools to estimate biological age and predict survival outcomes in patients with cancer.Biological age reflects how old a person appears, rather than their chronological age. The findings suggest that facial photographs could offer a non-invasive way to assess aspects of a patient’s health and help inform treatment decisions. However, the technology requires further validation before it can be used routinely in clinical practice.Facial images have emerged as a potential biomarker for health assessment, with studies suggesting that AI can extract health-related information from facial features. “As the population ages, physicians need better ways to estimate physical reserve and frailty before recommending intensive treatments such as radiation, surgery or systemic therapy,” said Dr Raymond Mak, a radiation oncologist at the Mass General Brigham Cancer Institute and a faculty member of its Artificial Intelligence in Medicine (AIM) programme. Mak was the study’s corresponding author.He said the aim was to make health assessments based on facial photographs more quantitative, consistent and scalable, enabling more personalised treatment decisions without replacing clinical judgement.Two AI tools assess biological age and survival prospectsThe researchers developed two tools using the FAHR-Face foundation model: FAHR-FaceAge, which estimates biological age, and FAHR-FaceSurvival, which directly predicts survival prospects from a facial photograph. The study found that the difference between a patient’s estimated biological age and chronological age was independently associated with survival. The researchers also reported that the models performed robustly across common variations in photographs.The findings build on earlier work by the same team, which published a model called FaceAge last year. The new FAHR-Face model was trained on a much larger collection of more than 40 million publicly available facial photographs, spanning a broader age range and a more diverse population.Story continues below this adHigher estimated biological age linked to greater mortality riskThe researchers evaluated the tools in more than 38,000 cancer patients treated with radiotherapy in the United States. Patients whose estimated biological age was at least five years higher than their chronological age had a 21 per cent higher risk of death during an average follow-up of four years than other patients. Those whose estimated biological age was at least five years lower had a 17 per cent lower risk.FAHR-FaceSurvival assigned nearly 4,000 patients a risk score ranging from zero to one. Among patients with scores between 0.9 and 1, indicating the highest predicted risk, 53 per cent (64 of 122) died within 90 days. By comparison, 2 per cent (20 of 975) of those with scores between zero and 0.1 died during the same period.The researchers found that combining the two models predicted survival more accurately than using either model alone. “The key technical advance is that FAHR-Face learns a general representation of the face from more than 40 million images, which can then be adapted to distinct health outcomes such as biological age and survival,” said Hugo Aerts, co-senior author of the study and director of the AIM programme at Mass General Brigham.The two tools also analyse different facial regions. FAHR-FaceAge focuses on features including the nasolabial folds, forehead and temples, while FAHR-FaceSurvival evaluates areas below the eyes and around the nasal bridge.Story continues below this ad“To our knowledge, this is the first demonstration that an AI model can directly predict cancer survival from a simple photograph,” Aerts said.Smoking linked to older-looking facesThe researchers also examined associations between lifestyle factors and estimated facial ageing. Heavy alcohol use, drug use and smoking were associated with an older estimated biological age.Smoking showed the strongest pattern: current smokers tended to look older than former smokers, who, in turn, tended to look older than people who had never smoked.The findings suggest that a facial photograph could provide a relatively simple and low-cost source of information to complement existing clinical assessments. However, the researchers said further studies are needed to establish the models’ reliability and determine how they might be used in clinical decision-making. The team is continuing its work through the FaceAge Research Study, an online study approved by an institutional review board and open to the public at faceage.bwh.harvard.edu.Story continues below this adParticipants can upload a facial photograph and receive a FaceAge research readout. The data will also help researchers refine and validate AI-based measures of facial ageing and health across larger and more diverse populations.