Endometrial whole-slide images dataset for detection of malignancy in endometrial biopsies

Mahnaz Mohammadi; Christina Fell; Sarah Bell; Gareth Bryson; Sheeba Syed; Prakash Konanahalli; David Harris-Birtill; Ognjen Arandjelovic; Clare Orange; Prishma Shahi; In Hwa Um; James D Blackwood; David J Harrison

doi:10.1093/gigascience/giaf147

Endometrial whole-slide images dataset for detection of malignancy in endometrial biopsies

Mahnaz Mohammadi^*, Christina Fell, Sarah Bell, Gareth Bryson, Sheeba Syed, Prakash Konanahalli, David Harris-Birtill, Ognjen Arandjelovic, Clare Orange, Prishma Shahi, In Hwa Um^*, James D Blackwood, David J Harrison

^*Corresponding author for this work

Research output: Contribution to journal › Article › peer-review

Abstract

Background: Whole-slide imaging enables the digitization of entire histological slides at a high resolution, allowing pathologists and
researchers to analyze tissue samples digitally rather than through traditional microscopy. This technology has become increasingly
valuable in pathology for research, education, and clinical diagnostics. Endometrial biopsy is very common, often being undertaken
to exclude noncancerous disease. This means that most cases do not contain cancer, and the challenge is to accurately and efficiently exclude serious pathology rather than simply make a diagnosis of malignancy. A well-curated, expert-annotated, endometrial whole-slide dataset covering a spread of cancer and noncancer diagnoses will support machine learning applications in automated diagnosis, facilitate research into the pathology of endometrial cancer, and serve as an educational resource for medical professionals.

Results: We introduce a newly constructed, large-scale dataset of endometrial biopsy specimens, comprising 2,909 whole-slide images
in iSyntax format, each accompanied by a corresponding annotation file in JSON format. Each whole-slide image is labeled with a primary class label representing its final diagnosis and a subcategory label providing further details within that diagnostic class.
These class labels are critical for machine learning applications, as they enable the development of artificial intelligence models capable of distinguishing between different types of endometrial abnormalities, improving automated classification, and guiding clinical decision-making.

Conclusions: Constructing and curating a high-quality endometrial whole-slide dataset requires significant effort to ensure accurate
annotations, data integrity, and patient privacy protection. However, the availability of a well-annotated dataset with detailed class labels is crucial for advancing digital pathology. Such a resource can enhance diagnostic accuracy, support personalized treatment strategies, and ultimately improve outcomes for patients with endometrial cancer and other endometrial conditions.

Original language	English
Article number	giaf147
Pages (from-to)	1-8
Number of pages	8
Journal	GigaScience
Volume	14
Early online date	5 Dec 2025
DOIs	https://doi.org/10.1093/gigascience/giaf147
Publication status	Published - 2025

Keywords

Endometrium
Whole-slide imaging
Endometrial cancer
Endometrial hyperplasia
Endometrial carcinoma
Digital slide repository
Image analysis
Image-segmentation
Histopathology
Deep learning
Machine learning

Access to Document

10.1093/gigascience/giaf147Licence: CC BY

Mohammadi-et-al-2025-Endometrial-whole-slide-GigaSci-giaf147-CCBY
Copyright © The Author(s) 2025. Published by Oxford University Press on behalf of GigaScience. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted reuse, distribution, and reproduction in any medium, provided the original work is properly cited.
Final published version, 3.68 MBLicence: CC BY

https://academic.oup.com/gigascience/article/doi/10.1093/gigascience/giaf147/8371767Licence: CC BY

ICAIRD: I-CAIRD: Industrial Centre for AI Research in Digital Diagnostics
Harris-Birtill, D. (PI) & Arandelovic, O. (CoI)
Technology Strategy Board
1/02/19 → 31/01/23
Project: Standard
ICAIRD: I-CAIRD: Industrial Centre for AI Research in Digital Diagnostics
Harrison, D. (PI)
Innovate UK
1/02/19 → 31/01/22
Project: Standard

Endometrial Whole Slide Images Dataset
Um, I. (Creator), Mohammadi, M. (Creator), Fell, C. (Creator), Morrison, D. (Creator), Orange, C. E. L. (Creator), Harrison, D. (Creator), Harris-Birtill, D. (Creator), Arandelovic, O. (Creator) & Blackwood, J. (Creator), EMBL-EBI, 2024
DOI: 10.6019/S-BIAD1199
Dataset

Cite this

@article{218e861562394f51ac09f17d73ee29a0,

title = "Endometrial whole-slide images dataset for detection of malignancy in endometrial biopsies",

abstract = "Background: Whole-slide imaging enables the digitization of entire histological slides at a high resolution, allowing pathologists and researchers to analyze tissue samples digitally rather than through traditional microscopy. This technology has become increasingly valuable in pathology for research, education, and clinical diagnostics. Endometrial biopsy is very common, often being undertaken to exclude noncancerous disease. This means that most cases do not contain cancer, and the challenge is to accurately and efficiently exclude serious pathology rather than simply make a diagnosis of malignancy. A well-curated, expert-annotated, endometrial whole-slide dataset covering a spread of cancer and noncancer diagnoses will support machine learning applications in automated diagnosis, facilitate research into the pathology of endometrial cancer, and serve as an educational resource for medical professionals. Results: We introduce a newly constructed, large-scale dataset of endometrial biopsy specimens, comprising 2,909 whole-slide images in iSyntax format, each accompanied by a corresponding annotation file in JSON format. Each whole-slide image is labeled with a primary class label representing its final diagnosis and a subcategory label providing further details within that diagnostic class. These class labels are critical for machine learning applications, as they enable the development of artificial intelligence models capable of distinguishing between different types of endometrial abnormalities, improving automated classification, and guiding clinical decision-making. Conclusions: Constructing and curating a high-quality endometrial whole-slide dataset requires significant effort to ensure accurate annotations, data integrity, and patient privacy protection. However, the availability of a well-annotated dataset with detailed class labels is crucial for advancing digital pathology. Such a resource can enhance diagnostic accuracy, support personalized treatment strategies, and ultimately improve outcomes for patients with endometrial cancer and other endometrial conditions.",

keywords = "Endometrium, Whole-slide imaging, Endometrial cancer, Endometrial hyperplasia, Endometrial carcinoma, Digital slide repository, Image analysis, Image-segmentation, Histopathology, Deep learning, Machine learning",

author = "Mahnaz Mohammadi and Christina Fell and Sarah Bell and Gareth Bryson and Sheeba Syed and Prakash Konanahalli and David Harris-Birtill and Ognjen Arandjelovic and Clare Orange and Prishma Shahi and Um, \{In Hwa\} and Blackwood, \{James D\} and Harrison, \{David J\}",

note = "Funding: This work is supported by the Industrial Centre for AI Research in digital Diagnostics (iCAIRD), which is funded by Innovate UK on behalf of UK Research and Innovation (UKRI) [project number: 104690] and in part by Chief Scientist Office, Scotland.",

year = "2025",

doi = "10.1093/gigascience/giaf147",

language = "English",

volume = "14",

pages = "1--8",

journal = "GigaScience",

issn = "2047-217X",

publisher = "Oxford University Press",

}

TY - JOUR

T1 - Endometrial whole-slide images dataset for detection of malignancy in endometrial biopsies

AU - Mohammadi, Mahnaz

AU - Fell, Christina

AU - Bell, Sarah

AU - Bryson, Gareth

AU - Syed, Sheeba

AU - Konanahalli, Prakash

AU - Harris-Birtill, David

AU - Arandjelovic, Ognjen

AU - Orange, Clare

AU - Shahi, Prishma

AU - Um, In Hwa

AU - Blackwood, James D

AU - Harrison, David J

N1 - Funding: This work is supported by the Industrial Centre for AI Research in digital Diagnostics (iCAIRD), which is funded by Innovate UK on behalf of UK Research and Innovation (UKRI) [project number: 104690] and in part by Chief Scientist Office, Scotland.

PY - 2025

Y1 - 2025

N2 - Background: Whole-slide imaging enables the digitization of entire histological slides at a high resolution, allowing pathologists and researchers to analyze tissue samples digitally rather than through traditional microscopy. This technology has become increasingly valuable in pathology for research, education, and clinical diagnostics. Endometrial biopsy is very common, often being undertaken to exclude noncancerous disease. This means that most cases do not contain cancer, and the challenge is to accurately and efficiently exclude serious pathology rather than simply make a diagnosis of malignancy. A well-curated, expert-annotated, endometrial whole-slide dataset covering a spread of cancer and noncancer diagnoses will support machine learning applications in automated diagnosis, facilitate research into the pathology of endometrial cancer, and serve as an educational resource for medical professionals. Results: We introduce a newly constructed, large-scale dataset of endometrial biopsy specimens, comprising 2,909 whole-slide images in iSyntax format, each accompanied by a corresponding annotation file in JSON format. Each whole-slide image is labeled with a primary class label representing its final diagnosis and a subcategory label providing further details within that diagnostic class. These class labels are critical for machine learning applications, as they enable the development of artificial intelligence models capable of distinguishing between different types of endometrial abnormalities, improving automated classification, and guiding clinical decision-making. Conclusions: Constructing and curating a high-quality endometrial whole-slide dataset requires significant effort to ensure accurate annotations, data integrity, and patient privacy protection. However, the availability of a well-annotated dataset with detailed class labels is crucial for advancing digital pathology. Such a resource can enhance diagnostic accuracy, support personalized treatment strategies, and ultimately improve outcomes for patients with endometrial cancer and other endometrial conditions.

AB - Background: Whole-slide imaging enables the digitization of entire histological slides at a high resolution, allowing pathologists and researchers to analyze tissue samples digitally rather than through traditional microscopy. This technology has become increasingly valuable in pathology for research, education, and clinical diagnostics. Endometrial biopsy is very common, often being undertaken to exclude noncancerous disease. This means that most cases do not contain cancer, and the challenge is to accurately and efficiently exclude serious pathology rather than simply make a diagnosis of malignancy. A well-curated, expert-annotated, endometrial whole-slide dataset covering a spread of cancer and noncancer diagnoses will support machine learning applications in automated diagnosis, facilitate research into the pathology of endometrial cancer, and serve as an educational resource for medical professionals. Results: We introduce a newly constructed, large-scale dataset of endometrial biopsy specimens, comprising 2,909 whole-slide images in iSyntax format, each accompanied by a corresponding annotation file in JSON format. Each whole-slide image is labeled with a primary class label representing its final diagnosis and a subcategory label providing further details within that diagnostic class. These class labels are critical for machine learning applications, as they enable the development of artificial intelligence models capable of distinguishing between different types of endometrial abnormalities, improving automated classification, and guiding clinical decision-making. Conclusions: Constructing and curating a high-quality endometrial whole-slide dataset requires significant effort to ensure accurate annotations, data integrity, and patient privacy protection. However, the availability of a well-annotated dataset with detailed class labels is crucial for advancing digital pathology. Such a resource can enhance diagnostic accuracy, support personalized treatment strategies, and ultimately improve outcomes for patients with endometrial cancer and other endometrial conditions.

KW - Endometrium

KW - Whole-slide imaging

KW - Endometrial cancer

KW - Endometrial hyperplasia

KW - Endometrial carcinoma

KW - Digital slide repository

KW - Image analysis

KW - Image-segmentation

KW - Histopathology

KW - Deep learning

KW - Machine learning

U2 - 10.1093/gigascience/giaf147

DO - 10.1093/gigascience/giaf147

M3 - Article

SN - 2047-217X

VL - 14

SP - 1

EP - 8

JO - GigaScience

JF - GigaScience

M1 - giaf147

ER -

Endometrial whole-slide images dataset for detection of malignancy in endometrial biopsies

Abstract

Keywords

Access to Document

Fingerprint

Projects

ICAIRD: I-CAIRD: Industrial Centre for AI Research in Digital Diagnostics

ICAIRD: I-CAIRD: Industrial Centre for AI Research in Digital Diagnostics

Datasets

Endometrial Whole Slide Images Dataset

Cite this