UniToPatho, a labeled histopathological dataset for colorectal polyps classification and adenoma dysplasia grading

Abstract

Histopathological characterization of colorectal polyps allows to tailor patients' management and follow up with the ultimate aim of avoiding or promptly detecting an invasive carcinoma. Colorectal polyps characterization relies on the histological analysis of tissue samples to determine the polyps malignancy and dysplasia grade. Deep neural networks achieve outstanding accuracy in medical patterns recognition, however they require large sets of annotated training images. We introduce UniToPatho, an annotated dataset of 9536 hematoxylin and eosin (H&E) stained patches extracted from 292 whole-slide images, meant for training deep neural networks for colorectal polyps classification and adenomas grading. We present our dataset and provide insights on how to tackle the problem of automatic colorectal polyps characterization.

Classification of Colorectal Polyps from Histopathological Images using Ensemble of ConvNeXt Variants

“…UniToPatho database contains 9536 hematoxylin and eosin stained patches extracted from 292 whole-slide images, each of the slides have a magnification of 20×. Moreover, the WSIs belong to the following classes; normal tissue, hyperplastic polyp, tubular adenoma and tubulo-villous adenoma [29]. EBHI is composed of 5532 WSIs which has the following categories, normal, low-grade and high-grade intraepithelial neoplasm, and adenocarcinoma, divided into four magnifications of 40×, 100×, 200× and 400× [30].…”

“…• In this study, we explored state-of-the-art pre-trained Deep CNN algorithms' performances on our custom dataset. In order to comprehensively evaluate and assess the generalizability of the proposed model, during the testing phase, we also employ publicly available UniToPatho and EBHI databases [29], [30]. The proposed ensemble model achieves an accuracy of 95% on our custom dataset.…”