A foundation model for breast and lung cancer screening using non-contrast computed tomography

Abstract Cancer screening can enable early detection and improve survival but a focus on single cancers limits cost-effectiveness. Here we present OMAFound (carcinOMA Finder foundation), a foundation model capable of simultaneous multi-cancer screening at both organ level and patient level using widely accessible non-contrast computed tomography (CT). The model was developed and tested on 325,197 CT volumes from 151,386 patients across 10 Chinese and international datasets, achieving performance comparable to mammography-based approaches for breast cancer detection and matching existing lung-specific models for lung cancer detection. In a prospective multi-centre cohort of 21,601 patients undergoing low-dose CT screening, OMAFound demonstrated balanced accuracy of 82.2% for breast cancer and 88.0% for lung cancer in females, while attaining 86.1% balanced accuracy for lung cancer detection in males. When assisted by OMAFound, 7 generalist radiologists showed improvement in sensitivity (mean increases of 38.9% for breast cancer, 16.0% for lung cancer and 21.3% at patient level), without compromising specificity. These findings highlight the potential of OMAFound as a multi-cancer screening tool to offer robust preventive medicine strategies with minimal costs.