BRACS: A Dataset for BReAst Carcinoma Subtyping in H&E Histology Images

TL;DR

BRACS is a large, annotated dataset of breast cancer histology images designed to facilitate AI research in tumor subtyping, including rare atypical lesions, to improve diagnostic accuracy and consistency.

Contribution

The paper introduces BRACS, the largest annotated dataset for breast cancer subtyping in histology images, including detailed annotations of benign, malignant, and atypical lesions.

Findings

Contains 547 WSIs and 4539 ROIs with expert annotations.

Includes diverse lesion subtypes, especially understudied atypical lesions.

Provides a valuable resource for developing AI diagnostic tools.

Abstract

Breast cancer is the most commonly diagnosed cancer and registers the highest number of deaths for women with cancer. Recent advancements in diagnostic activities combined with large-scale screening policies have significantly lowered the mortality rates for breast cancer patients. However, the manual inspection of tissue slides by the pathologists is cumbersome, time-consuming, and is subject to significant inter- and intra-observer variability. Recently, the advent of whole-slide scanning systems have empowered the rapid digitization of pathology slides, and enabled to develop digital workflows. These advances further enable to leverage Artificial Intelligence (AI) to assist, automate, and augment pathological diagnosis. But the AI techniques, especially Deep Learning (DL), require a large amount of high-quality annotated data to learn from. Constructing such task-specific datasets poses several challenges, such as, data-acquisition level constrains, time-consuming and expensive annotations, and anonymization of private information. In this paper, we introduce the BReAst Carcinoma Subtyping (BRACS) dataset, a large cohort of annotated Hematoxylin & Eosin (H&E)-stained images to facilitate the characterization of breast lesions. BRACS contains 547 Whole-Slide Images (WSIs), and 4539 Regions of Interest (ROIs) extracted from the WSIs. Each WSI, and respective ROIs, are annotated by the consensus of three board-certified pathologists into different lesion categories. Specifically, BRACS includes three lesion types, i.e., benign, malignant and atypical, which are further subtyped into seven categories. It is, to the best of our knowledge, the largest annotated dataset for breast cancer subtyping both at WSI- and ROI-level. Further, by including the understudied atypical lesions, BRACS offers an unique opportunity for leveraging AI to better understand their characteristics.