BRACS(BReAst Carcinoma Subtyping)

BReAst Carcinoma Subtyping (BRACS) 数据集，一大组带注释的苏木精和伊红 (H&E) 染色图像，以促进乳腺病变的表征。 BRACS 包含 547 个全幻灯片图像 (WSI) 和从 WSI 中提取的 4539 个感兴趣区域 (ROI)。每个 WSI 和各自的 ROI 都由三位经过董事会认证的病理学家共识注释为不同的病变类别。具体而言，BRACS 包括三种病变类型，即良性、恶性和非典型，进一步细分为七类。

The Breast Carcinoma Subtyping (BRACS) dataset is a large collection of annotated hematoxylin and eosin (H&E)-stained images developed to facilitate the characterization of breast lesions. The BRACS dataset comprises 547 whole-slide images (WSIs) and 4,539 regions of interest (ROIs) extracted from these WSIs. Each WSI and its corresponding ROIs were annotated by consensus of three board-certified pathologists into distinct lesion categories. Specifically, the BRACS dataset includes three major lesion types: benign, malignant, and atypical, which are further subdivided into seven categories.

搜集汇总

背景与挑战

背景概述

BRACS数据集是一个用于乳腺癌亚型分类的医疗图像数据集，包含547个全幻灯片图像和4539个感兴趣区域，基于H&E染色图像，由病理学家标注为三种病变类型（良性、恶性和非典型）并细分为七类。该数据集旨在促进乳腺病变的自动表征研究，发布于2021年，适用于计算机视觉和医疗诊断领域。