Built by Paige and Microsoft, PRISM2 reads whole-slide images through a perceiver-based encoder trained jointly on tissue tiles and clinical dialogue drawn from pathology reports. The model aggregates thousands of tile embeddings per slide into one representation, then generates text that…
Read Full Article at Source
