Facial photographs as proxies for inflammatory aging

Systemic chronic inflammation is a major determinant of aging and disease risk, yet current biomarkers such as the Inflammatory Age (iAge) clock and other proteomic assays remain costly, invasive, and poorly scalable. The skin, as both a visible marker and contributor to age-related inflammation (inflammaging), offers a potential non-invasive window into this health metric. Here, we developed Healthy Selfie, a digital predictor of iAge, from simple 2D frontal face photographs. We leveraged data from the Edifice Health Pre-Market trial, a clinical study of 750 participants aged 20 to 90 years, using a subset with available iAge measurements paired to facial images, demographic, clinical, and functional data. Facial embeddings from a pre-trained deep learning image model were combined with chronological age, sex and other easily obtainable metadata, and mapped to blood-derived iAge using regression within a leave-one-out cross-validation framework. Photo-predicted iAge values were significantly correlated with blood iAge values (r = 0.536) and were used to identify increased iAge acceleration (accuracy 66.39%, sensitivity 65.61%, specificity 67.24%). Validation on an external dataset containing ~100,000 images showed no significant demographic bias across ethnicities, including Asian, Black, Indian, Middle Eastern, and Latino populations. Beyond iAge, we also showed that facial features could predict blood levels of individual inflammaging proteins (CXCL9, CXCL1, CCL11, TNFSF10, and IFNG). Our findings suggest that ordinary facial photos can provide information on blood inflammation and can be used as a scalable, ultra low-cost tool for assessing biological aging and advancing precision health.