Placenta
收藏资源简介:
Placenta数据集是由牛津大学妇女与生殖健康系创建,用于在未充分探索的领域——胎盘组织学全切片图像中的细胞图预测微观解剖组织结构。该数据集包含来自两个机构的两个胎盘组织学图像中的两个细胞图,总计2,395,747个节点,其中799,745个节点具有真实标签。数据集的创建过程涉及使用深度学习管道识别全切片图像中所有细胞的中心点,并将其分类为11种细胞类型之一。Placenta数据集的应用领域主要集中在科学研究中,旨在通过自动化方法量化组织结构单元,未来可能应用于临床组织病理学诊断和患者护理,同时也为开发新型可扩展图神经网络提供基准,以适应节点特征和类别标签的不平衡,理解图中的大小节点社区,并能抵抗数据的不完整性。
The Placenta Dataset was developed by the Department of Women's and Reproductive Health, University of Oxford, for predicting micro-anatomical tissue structures from cell graphs in whole-slide histopathology images of the placenta, an under-explored research domain. This dataset includes two cell graphs derived from two placental histopathology images sourced from two distinct institutions, containing a total of 2,395,747 nodes, out of which 799,745 nodes are annotated with ground-truth labels. The construction of this dataset utilizes a deep learning pipeline to identify the centroid coordinates of every cell within the whole-slide images and classify each cell into one of 11 cell types. The Placenta Dataset is primarily applied in scientific research, aiming to quantify tissue structural units through automated methods. In the future, it may be utilized in clinical histopathological diagnosis and patient care. Additionally, it serves as a benchmark for developing novel scalable graph neural networks, enabling adaptation to the imbalance of node features and category labels, understanding node communities of varying scales in graphs, and resisting data incompleteness.




