Welcome to the Cloud Wars Minute — your daily cloud news and commentary show. Each episode provides insights and perspectives around the “reimagination machine” that is the cloud.
In today’s Cloud Wars Minute, I look at how Microsoft Research and Paige are rethinking pathology AI by training PRISM 2 on images and real-world clinical language.
Highlights
00:04 — Microsoft Research and the healthcare company Paige have developed a pathology foundation model called PRISM 2. Now, what makes it interesting is the model is trained on both tissue images and real-world pathology report language.
00:31 — The difference is that many traditional AI systems are built around a specific task. PRISM 2, on the other hand, is designed to handle multiple cancer detection, biomarker prediction, and research tasks without having to build a completely new model every time. The researchers report that PRISM 2 performs strongly across a range of pathology benchmarks.
Community Summit North America is the largest independent innovation, education, and training event for Microsoft business applications delivered by Expert Users, Microsoft Leaders, MVPs, and Partners. Register now to attend Community Summit in Nashville, TN from October 11-15.
00:54 — The model has also been made available for non-commercial research, which could really help push the field forward further. What I find really interesting is that PRISM 2 is trained on almost two and a half million whole-slide images, along with the pathology reports that go with them.
01:10 — The model isn’t just learning what cancer looks like under a microscope, but is also learning how pathologists actually describe what they can see. Advancing healthcare has always been one of the core goals of AI, but for that to happen, I think it’s becoming increasingly clear that we need to think differently about how we train these models.
01:36 — I think this is a really interesting step forward, specifically in pathology, and with the model available for research, it could give researchers another tool to explore what’s possible when you bring medical images and clinical language together.
