Multi-Laboratory Hematoxylin and Eosin Staining Variance Unsupervised Machine Learning Dataset

We provide the generated dataset used for unsupervised machine learning in [1]. The data is in CSV format and contains all principal components and ground truth labels, per tissue type. Tissue type codes used are; C1 for kidney, C2 for skin, C3 for colon, and 'PC' for the principal component. Please see the original design in [1] for feature extraction specifications. Features have been extracted independently for each tissue type.

Reference:
Prezja, F.; Pölönen, I.; Äyrämö, S.; Ruusuvuori, P.; Kuopio, T. H&E Multi-Laboratory Staining Variance Exploration with Machine Learning. Appl. Sci. 2022, 12, 7511. https://doi.org/10.3390/app12157511

本研究提供了文献[1]中用于无监督机器学习的生成数据集。该数据集采用CSV格式存储，包含所有主成分（principal components）以及对应每种组织类型的真实标签（ground truth labels）。所用组织类型编码如下：C1代表肾脏，C2代表皮肤，C3代表结肠，'PC'代表主成分（principal component）。有关特征提取的具体规范，请参阅文献[1]中的原始设计方案。特征均针对各组织类型独立提取。

参考文献：
Prezja, F.; Pölönen, I.; Äyrämö, S.; Ruusuvuori, P.; Kuopio, T. 基于机器学习的H&E染色多实验室差异探究. 应用科学, 2022, 12, 7511. https://doi.org/10.3390/app12157511