CRC-HGD-v1: A Histopathological Image Dataset for Grading Colorectal Cancer

Published: 15 September 2026| Version 5 | DOI: 10.17632/yfp5sfj47m.5
Contributors:
Elham Amjadi, Amin Bahreini, Sayyed Mohammadreza Hakimian, Mohammad Hasan Emami, Alireza Fahim, Hojjatollah Rahimi,

Description

CRC-HGD-v1 is the first version of the Colorectal Cancer Histopathological Grading Dataset. This dataset includes 1,914 images in five categories: (1) Grade I — Well Differentiated, (2) Grade II — Moderately Differentiated, (3) Grade III — Poorly Differentiated, (4) Normal colorectal tissue, and (5) Mixed Normal/Tumoral tissue. The three main tumor grade categories (Grade I, II, and III) consist of images captured at four magnification levels: 4x, 10x, 20x, and 40x. For each specimen, one image is provided at 4x magnification; the number of images at higher magnification levels (10x, 20x, 40x) may vary per specimen based on the pathologist's assessment, with additional images captured when clinically informative. Overall, the dataset contains 106 patients in Grade I, 75 patients in Grade II, and 33 patients in Grade III, totaling 214 patients. Considering all images across the four magnification levels, there are 860 images in Grade I, 712 in Grade II, and 327 in Grade III. The dataset also includes 8 images of normal colorectal tissue (2 images per magnification level) and 7 mixed normal-tumoral images. All specimens are colorectal tissue sections stained with Hematoxylin and Eosin (H&E), collected between 2014 and 2019. Histologic grading was performed according to the World Health Organization (WHO) criteria, based on the proportion of gland-forming structures: Grade I (well-differentiated, >95% gland formation), Grade II (moderately differentiated, 50–95% gland formation), and Grade III (poorly differentiated, <50% gland formation). These samples were gathered and prepared by a multidisciplinary team of medical and computational experts in pathology, surgery, gastroenterology, artificial intelligence (AI), and data science, under the supervision of the Poursina Hakim Digestive Disease Research Center (PDDRC), Isfahan University of Medical Sciences, Isfahan, Iran. Image specifications: JPEG format, RGB color space, 800 × 800 pixels resolution, 96 DPI. File naming convention: [SampleNumber]_CRC_[GradeLabel]_[ArchiveNumber]_[Magnification]_[ImageIndex] Example: 01_CRC_G1_16130_4x_1 ** For the latest updates and more information about this dataset, please refer to: https://databiox.com

Files

Institutions

Categories

Artificial Intelligence, Computer Vision, Pathology, Medical Imaging, Colorectal Medicine, Histopathology, Colorectal Cancer, Colorectal Surgery, Convolutional Neural Network, Deep Learning, Image Analysis

Licence