PRISM: A Multi-Modal Generative Foundation Model for Slide-Level Histopathology
Published in arXiv preprint, 2024
Abstract: We introduce PRISM, a multi-modal generative foundation model for slide-level histopathology. Rather than processing individual image tiles separately, PRISM operates at the whole-slide level using Virchow tile embeddings and clinical report text. The model demonstrates zero-shot cancer detection and sub-typing with performance comparable to supervised methods, while also showing strong results in biomarker prediction tasks with limited training data.
Recommended citation: G Shaikovski et al. (2024). "PRISM: A Multi-Modal Generative Foundation Model for Slide-Level Histopathology" arXiv:2405.10254. https://arxiv.org/abs/2405.10254