Evaluating Multimodal Pathology Foundation Models for Clinical Readiness Assessment: A Systematic Benchmark for Region-of-Interest Pathology Across Organs, Tasks and Supervision Regimes

In this paper, we release a comprehensive pediatric cohort encompassing five distinct sub-datasets and propose a large-scale systematic benchmark. The curated pediatric cohort comprises five distinct astrocytoma sub-datasets: Pilocytic Astrocytoma (PA) (7,812 patches, 8 classes), Pilomyxoid Astrocytoma (PMA) (16,512 patches, 9 classes), Desmoplastic Infantile Astrocytoma (DIA) (20,813 patches, 5 classes), Pleomorphic Xanthoastrocytoma (PXA) (19,737 patches, 10 classes), and Subependymal Giant Cell Astrocytoma (SGCA) (42,885 patches, 10 classes).