The abstract is read at the publisher; the summary is JClub's.
CROWN, a DINOv2-based self-supervised visual foundation model, achieves over 95% accuracy in 48 patch-level classification evaluations and 98% in 22.
- Why it matters: Cytopathology is crucial for cancer screening and diagnosis but is labor-intensive, creating demand for scalable models that generalize across tasks and datasets.
- What they did: CROWN was pretrained on more than 10 million cytology image patches without manual annotations, then evaluated across 202 task-setting combinations spanning multiple vision tasks.
- The result: Across private institutional cohorts and public datasets, CROWN delivered the best overall performance among established pretrained encoders, supporting it as a general-purpose backbone for heterogeneous computational cytopathology.
The findingWhy it mattersWhat they didThe result
AI summary of the abstract · may contain errors · not medical advice · read the paper.