DAP Atlas
收藏资源简介:
DAP Atlas数据集是由卡尔斯鲁厄理工学院人类学与机器人学研究所开发的一个全面的全身CT图像数据集。该数据集包含533个CT图像,每个图像有142个体素级标签,覆盖了从身体组成到各种血管的广泛解剖信息。数据集的创建过程涉及使用nnU-Net基础的伪标签生成和解剖学指导的伪标签细化,不依赖于手动标注。DAP Atlas数据集旨在解决当前医学数据模型主要专注于部分标注数据集的问题,通过提供一个全面的解剖视图,支持临床相关的下游任务,如身体成分分析、手术规划或癌症治疗监测。
The DAP Atlas is a comprehensive whole-body CT image dataset developed by the Institute of Anthropology and Robotics, Karlsruhe Institute of Technology (KIT). It contains 533 CT images, each paired with 142 voxel-level labels, covering a wide range of anatomical information ranging from body composition to various blood vessels. The dataset development process employs pseudo-label generation based on nnU-Net and anatomy-guided pseudo-label refinement, without relying on manual annotations. The DAP Atlas aims to address the issue that current medical data models primarily focus on partially annotated datasets, by providing a comprehensive anatomical view to support clinically relevant downstream tasks such as body composition analysis, surgical planning, or cancer treatment monitoring.




