fadefade/OmniTumorData

OmniTumorData是一个精心策划的多源基准数据集，用于跨CT和MRI的文本提示3D肿瘤分割。该数据集不直接分发任何影像数据，而是提供数据集清单、21提示本体、147字符串同义增强池、案例级训练/验证/测试分割以及PNG转换脚本。数据集由12个队列组成，共11,326个受试者，涵盖多种解剖结构和模态。数据集适用于图像分割任务，并包含详细的预处理步骤和引用要求。

OmniTumorData is a curated multi-source benchmark for text-prompted 3D tumor segmentation across CT and MRI. The repository does not redistribute any imaging data but provides a dataset inventory with download links, a 21-prompt ontology, a 147-string synonymous augmentation pool, case-level train/val/test splits, and PNG conversion scripts. The dataset is composed of 12 cohorts with a total of 11,326 subjects, covering various anatomies and modalities. It is designed for image-segmentation tasks and includes detailed preprocessing steps and citation requirements.