A universal visual foundation model for computational cytopathology.

CROWN achieves over 95% accuracy in 48 cytology patch-level classifications, establishing it as a highly effective universal model for computational cytopathology.

• Why it matters: Cytopathology is vital for cancer diagnosis but remains labor-intensive, highlighting the need for scalable, automated computational tools to improve efficiency and consistency.
• What they did: Researchers pretrained CROWN on more than 10 million cytology image patches using a self-supervised DINOv2-based framework, avoiding manual annotations, and evaluated it across 201 diverse tasks and datasets.
• The result: CROWN outperformed established pretrained encoders, with accuracy exceeding 98% in 22 evaluations, supporting its role as a versatile, scalable backbone for various cytopathology applications.