RadGenome-Anatomy is a large-scale chest radiograph anatomy segmentation dataset
constructed from the RadGenome-ChestCT corpus
(originally based on CT-RATE).
It contains 25,692 volumetric studies (24,128 train / 1,564 validation), yielding paired
postero-anterior (PA) and lateral (LL) projection images at 384 × 384 resolution.
Across the two radiographic views, the dataset provides 10,790,646 fine-grained anatomy masks
over 210 canonical anatomy classes and 513… See the full description on the dataset page:
https://huggingface.co/datasets/EvidenceAIResearch/radgenome-anatomy.