Researchers at the University of California, Berkeley, and the University of California, San Francisco have developed Pillar-0, an open-source AI model for analyzing medical images that demonstrates superior diagnostic accuracy. Achieving an average AUC of .87 across over 350 findings, Pillar-0 outperforms existing leading AI models from Google, Microsoft, and Alibaba by 10% to 17%. Unlike previous models limited to 2D images, Pillar-0 interprets 3D volumes directly, allowing it to recognize numerous conditions from a single CT or MRI scan. This model is expected to bridge the capacity gap for radiologists, given the high volume of annual imaging procedures. The research team has made Pillar-0 available to facilitate further developments in AI-powered medical imaging, aiming to improve diagnostic workflows and support hospitals in utilizing the technology effectively.
Wed, 10 Dec 2025 08:00:00 GMT | Imaging Technology News