SlideCheck: Guiding Self-Supervised Pretraining of Pathology Foundation Models via Dataset Distributions
Researchers have developed SlideCheck, a tool designed to guide the self-supervised pretraining of pathology foundation models. This tool operates by analyzing dataset distributions and providing explicit scores for abnormality and malignancy within Whole Slide Images (WSIs). By organizing, filtering, and auditing pretraining data based on these scores, SlideCheck aims to improve the efficiency and controllability of pathology foundation model development. Experiments demonstrate that data curated using SlideCheck can influence downstream model behavior and achieve performance comparable to models trained on full datasets. AI
IMPACT Enhances auditable and efficient pretraining data construction for pathology AI models.