Researchers at Mass General Brigham have made their artificial intelligence (AI) tool, FaceAge, publicly available, allowing users to estimate their biological age from facial photographs. The tool, initially developed for hospital patients, has now been opened to anyone willing to participate in the FaceAge Research Study, accessible online. Since its quiet launch earlier this year, over 10,000 individuals have enrolled.
Participants who join the study provide consent, complete a survey detailing their chronological age, sleep and exercise habits, and medical history, then upload three to ten selfies. FaceAge analyzes these images and returns an estimate of biological age—how old the individual's body appears based on facial features—within seconds. Users can continue to upload images over time to monitor changes.
The model behind FaceAge is based on a foundation AI called FAHR-Face, which was trained on more than 40 million facial images. Researchers subsequently fine-tuned it for two main purposes: estimating biological age and predicting cancer survival outcomes through a sub-model named FaceSurvival. A study published in a leading medical journal demonstrated that among 38,000 cancer patients at Harvard-affiliated hospitals, those whose faces appeared at least five years older than their chronological age faced a 21% higher risk of death, after adjusting for variables such as actual age, sex, and cancer type.
Hugo Aerts, director of the Artificial Intelligence in Medicine program at Mass General Brigham, highlighted this as the first known example of an AI model directly predicting cancer survival using simple photographs. However, external experts caution that while the approach is innovative, it relies on retrospective data rather than prospective clinical trials. One expert noted that the model’s predictive performance varied across different cancer types, showing significant associations in 18 of 28 categories evaluated.
FaceAge is currently being evaluated in two clinical trials. One study at Brigham and Women’s Hospital is enrolling 60 patients aged 65 and older with early-stage lung cancer before radiation therapy. The goal is to determine whether FaceAge can identify frail patients who may benefit from additional support such as nutrition or physical therapy ahead of treatment. Radiation oncologist Dr. Raymond Mak emphasized the potential of this tool to address the widespread underassessment of geriatric vulnerabilities among older lung cancer patients.
The second trial, led by thoracic surgeons at Mass General Brigham, combines FaceAge with data from wearable devices to predict patients at higher risk of surgical complications before major operations. This effort aims to enhance the assessment of physiologic reserve beyond conventional clinical measures.
Researchers used attention mapping techniques to gain insight into the facial regions influencing the AI's predictions. FaceAge predominantly focuses on the nasolabial folds and the forehead, while FaceSurvival looks mainly at areas beneath the eyes and over the nose bridge. However, the specific visual traits driving the model's estimates—such as wrinkles, pigmentation, or muscle wasting—remain unclear.
Initial concerns that variables like makeup, lighting, facial expression, or cosmetic surgery could skew results have been largely allayed. Most factors had minimal impact, though rhinoplasty appeared to slightly increase the estimated biological age, potentially due to tissue loss. Lifestyle factors showed clearer effects: heavy alcohol consumption, drug use, and especially smoking were linked to older biological appearance, with current smokers looking older than former smokers and non-smokers.
As the FaceAge database expands and researchers collaborate with external institutions, the tool’s algorithms are expected to improve across more diverse populations and health conditions. The developers stress that FaceAge is intended as a research tool rather than a diagnostic device, but ongoing clinical trials will clarify its potential applications in healthcare.