Histology Tissue Fold Dataset with Pixel-Level Annotations for Artificial Intelligence-Based Detection and Segmentation of Tissue Fold Artifacts in H&E-Stained Teaching Slides

Description

This dataset contains 2,127 hematoxylin and eosin (H&E)-stained histological images collected from histology teaching slides representing five different tissues: brain, kidney, liver, small intestine, and testis. Images were acquired at 10× magnification.

The dataset consists of 899 clean (artifact-free) images and 1,228 images containing tissue fold artifacts. Pixel-level binary segmentation masks are provided for all 1,228 tissue fold images, enabling the dataset to be used for both image classification and semantic segmentation tasks.

Images are organized according to tissue type and class, while the corresponding annotation masks are organized by tissue type. Metadata describing the images are also provided.

The dataset was developed to support research in artificial intelligence, computer vision, computational histology, and digital pathology, particularly for the automated detection and localization of tissue fold artifacts in H&E-stained histological images.

Slide-level image mapping

The file slide_image_mapping.xlsx provides the slide-level mapping for all images included in the dataset. Each image filename is linked to its corresponding histological slide ID, together with the tissue type and class label. This metadata allows images originating from the same histological slide to be identified and grouped consistently for slide-level data splitting and reproducible model evaluation.

Associated Publication

The dataset was developed and evaluated as part of the following peer-reviewed research article:

Koparir, O.F.; Tarakci Gencer, B.; Sengur, A.
Deep Learning-Assisted Quality Control of Histology Teaching Slides: Detection and Localization of Tissue Fold Artifacts in H&E-Stained Images.
Bioengineering 2026, 13, 937.
https://doi.org/10.3390/bioengineering13080937

Article page:
https://www.mdpi.com/2306-5354/13/8/937

The associated study evaluates deep learning approaches for the classification and pixel-level localization of tissue fold artifacts, together with explainability analyses, using this openly available dataset.

Citation

Researchers using the images, annotation masks, or metadata from this repository are kindly requested to cite both the dataset and the associated research article.

Dataset:

Koparir, O.F.; Tarakci Gencer, B.; Sengur, A. (2026). Histology Tissue Fold Dataset with Pixel-Level Annotations for Artificial Intelligence-Based Detection and Segmentation of Tissue Fold Artifacts in H&E-Stained Teaching Slides. Zenodo. https://doi.org/10.5281/zenodo.21493259

Associated article:

Koparir, O.F.; Tarakci Gencer, B.; Sengur, A. Deep Learning-Assisted Quality Control of Histology Teaching Slides: Detection and Localization of Tissue Fold Artifacts in H&E-Stained Images. Bioengineering 2026, 13, 937.
https://doi.org/10.3390/bioengineering13080937