Data for glomeruli characterization in histopathological images

The data presented hereis part of the whole slide imaging (WSI) datasets generated in European project AIDPATH. This data is also related to the research paper entitle “Glomerulosclerosis Identification in Whole Slide Images using Semantic Segmentation”, published in Computer Methods and Programs in Biomedicine Journal (DOI: 10.1016/j.cmpb.2019.105273) . In that article, different methods based on deep learning for glomeruli segmentation and their classification into normal and sclerotic glomerulous are presented and discussed. These data will encourage research on artificial intelligence (AI) methods, create and compare fresh algorithms, and measure their usability in quantitative nephropathology. Parameters for data collection: Tissue samples were collected with a biopsy needle having an outer diameter between 100μm and 300μm. Afterwards, paraffin blocks were prepared using tissue sections of 4μm and stained using Periodic acid–Schiff (PAS). Then, images at 20x magnification were selected. Description of data collection: The tissue samples were scanned at 20x with a Leica Aperio ScanScope CS scanner. Data format: DATASET_A_DIB: Raw data, original images in SVS format. DATASET_B_DIB: Classified: Detected glomeruli to be used for classification in PNG format. The data is composed of two datasets: 1.) DATASET_A: Raw data with 31 whole slide images (WSI) in SVS format. The size of the WSI range between 21651x10498 pixels and 49799 x 32359 pixles acquired at 20x. The images contain different types of glomeruli that were detected using the algorithms explained at the following article [https://doi.org/10.1016/j.cmpb.2019.105273]. The detected glomeruli are provided in DATASET_B. 2.) DATASET_B: 2,340 images with a single glomerulous, 1,170 normal glomeruli and 1,170 sclerosed glomeruli. All of them are in PNG format. Value of the Data · These data can be used for benchmarking to encourage further research on AI methods applied to digital pathology in nephrology. · The additional value of this data is that it has been acquired and evaluated by expert pathologists from different European countries. · All researches in digital pathology can benefit from these data, to test classification algorithms. And particularly for glomeruli identification in nephrology studies. · This data can be used for further development and new experiments in glomeruli classification with more classes, like focal glomeruli besides normal and sclerotic glomeruli.本数据集为欧洲AIDPATH项目所生成的全视野数字切片成像（Whole Slide Imaging, WSI）数据集的一部分。本数据集同时关联一篇发表于《Computer Methods and Programs in Biomedicine》期刊（DOI: 10.1016/j.cmpb.2019.105273）、题为《基于语义分割的全视野数字切片成像中肾小球硬化识别》的研究论文。该论文中介绍并讨论了多种基于深度学习的肾小球分割方法，以及将肾小球划分为正常与硬化型的分类方案。本数据集将助力人工智能（Artificial Intelligence, AI）方法相关研究，支持新型算法的开发与对比，并可用于评估这些方法在定量肾脏病理学中的可用性。 数据采集参数：采用外径介于100μm至300μm之间的活检针采集组织样本；随后以4μm厚度的组织切片制备石蜡包埋块，并使用过碘酸-雪夫（Periodic acid–Schiff, PAS）染色法进行染色；最终选取20倍放大倍率的图像。 数据采集说明：组织样本采用徕卡Aperio ScanScope CS扫描仪以20倍倍率进行扫描。 数据格式： DATASET_A_DIB：原始数据，为SVS格式的原始图像。 DATASET_B_DIB：分类数据，为用于分类任务的已检测肾小球图像，格式为PNG。 本数据集包含两个子数据集： 1. DATASET_A：原始数据，包含31份SVS格式的全视野数字切片成像（Whole Slide Imaging, WSI）图像。这些WSI的分辨率介于21651×10498像素至49799×32359像素之间，均为20倍倍率采集所得。图像中包含多种类型的肾小球，其检测采用上述论文[https://doi.org/10.1016/j.cmpb.2019.105273]中所述的算法实现，检测得到的肾小球已收录于DATASET_B中。 2. DATASET_B：包含2340张单肾小球图像，其中1170张为正常肾小球，1170张为硬化肾小球，所有图像均为PNG格式。 数据集价值： · 本数据集可作为基准测试集，助力肾脏病理学领域数字病理学相关人工智能方法的后续研究。 · 本数据集的额外价值在于，其采集与评估工作均由来自欧洲多国的病理专家完成。 · 所有数字病理学相关研究均可借助本数据集测试分类算法，尤其适用于肾脏病理学研究中的肾小球识别任务。 · 本数据集可用于开发更多分类类别的肾小球分类模型，例如在正常与硬化肾小球之外增加局灶性肾小球分类的相关实验。