DigitalShadow: An Early Warning System of Coronary Artery Disease based on Facial Foundation Model

Coronary artery disease (CAD) is responsible for approximately 17.8 million deaths each year, making it a leading cause of mortality among the elderly and a growing burden on healthcare systems, particularly in the context of global population aging. Although widespread, CAD is largely preventable, underscoring the need for early detection and proactive intervention. In this study, we introduce DigitalShadow, the first non-invasive early warning system for CAD detection powered by a fine-tuned facial foundation model. Pre-trained on 21 million facial images, the model is further adapted into LiveCAD using a dataset of 7,004 facial images from 1,751 patients across four hospitals in China. DigitalShadow functions passively and contactlessly, extracting facial features from live video streams without requiring user engagement, and achieves an AUC-ROC of 0.87. Integrated with a personalized health database, the system generates natural language risk reports and individualized health recommendations. With privacy as a central design principle, DigitalShadow supports secure, on-device deployment to ensure responsible and confidential handling of user data.