Histopathology pairs of hematoxylin-eosin and Ki67 stainings of testicular seminoma

About

Dataset contains 77 pairs of stained sections of testicular seminoma. Each pair is produced from the neighbouring sections of the tissue to guarantee the most similarity possible of the stained tissue. HE and Ki67 images are registered and aligned.

When used, please cite the following work: Petríková, D.; Cimrák, I.; Tobiášová, K. and Plank, L. (2024), Dataset of Hematoxylin-eosin and Ki67 histopathological image pairs complemented by algorithms providing Ki67 index, DATA, submitted.

Data composition:

Dataset is comprised of 154 PNG files. 77 files with prefix id_HE contain images of HE staining, and 77 files with prefix id_Ki67 contain images of Ki67 staining. Both files in each pair id_HE_string.png and id_Ki67_string.png have the same resolution with avarage value 35000px x 35000px. Data is available in 39 ZIP files each containig two HE. images and two of their Ki67 counterparts. For example, 03.zip contains the following files:

05_HE_slide-011-A1-S1-B11.png,

05_Ki67_slide-012-A1-S2-B12.png,

06_HE_slide-013-A1-S1-B13.png,

06_Ki67_slide-014-A1-S2-B14.png.

Image names contain an additional random string.  An example of a HE - Ki67 pair is in file  HE-Ki67_pair.png.

Data preprocessing:

Original scans of tissues contained pairs of HE and Ki67 staining with rotated and shifted tissues. Therefore a preprocessing of image registration was performed to find an affine transformation for rotation and translation. The concrete steps of image registration was described in [Petrikova 2024]. Provided images are already aligned.

Data usage:

This dataset is aimed for AI models to predict the Ki67 index or even to generate Ki67 staining. Since one pair of HE and Ki67 images contain two physically different sections of tissue (although the neighbouring), there is no one-to-one correspondence on cellural level. However patches, created on the same locations from the images in one pair, containe similar quantitative characteristics, such as number of cells, number of Ki67 positive cells, averaged cell size, etc. In [Petrikova2024B] we elaborate the usage of the dataset by computing Ki67 index of HE patches evaluated from corresponding Ki67 patch.    

References:

[Petrikova2024] Petríková, D.; Cimrák, I.; Tobiášová, K. and Plank, L. (2024). Ki67 Expression Classification from HE Images with Semi-Automated Computer-Generated Annotations. In Proceedings of the 17th International Joint Conference on Biomedical Engineering Systems and Technologies - BIOINFORMATICS; ISBN 978-989-758-688-0; ISSN 2184-4305, SciTePress, pages 536-544. DOI: 10.5220/0012535900003657

[Petrikova2024B] Petríková, D.; Cimrák, I.; Tobiášová, K. and Plank, L. (2024), Dataset of Hematoxylin-eosin and Ki67 histopathological image pairs complemented by algorithms providing Ki67 index, DATA, submitted.