WSSS4LUAD

组织病理学幻灯片是癌症诊断的金标准。它提供了关于肿瘤微环境 (TME) 的大量信息，不仅在解释肿瘤的发生和发展中起着至关重要的作用，而且还影响癌症患者的治疗效果和预后。不同类型组织之间的串扰与肿瘤进展高度相关。因此，迫切需要对不同组织进行分割和区分，以进行进一步的临床研究。

肺癌是全球癌症死亡的主要原因。在此挑战中，我们旨在对肺腺癌的H & E染色全幻灯片图像 (WSI) 进行组织语义分割。当前的挑战是，获得组织语义分割的像素级注释非常困难且耗时。受计算机视觉中的弱监督语义分割 (wss) 的启发，我们决定仅提供图像级注释来执行组织语义分割。

在这项挑战中，我们扫描了广东省人民医院 (GDPH) 的67个H & E染色载玻片，并从癌症基因组图谱 (TCGA) 中收集了20个WSIs。每位患者仅提取一个WSI。这项挑战的目标是仅使用图像级注释来实现三种常见且有意义的组织类型，即肿瘤上皮组织，肿瘤相关基质组织和正常组织的像素级预测。参与者仅获得用于机器学习算法训练的图像级注释 (3位多类标签)，以及用于验证和测试的像素级地面真相。

Histopathological slides are the gold standard for cancer diagnosis. They provide extensive information on the Tumor Microenvironment (TME), which not only plays a critical role in elucidating tumorigenesis and progression, but also influences the treatment efficacy and prognosis of cancer patients. Crosstalk between different tissue types is highly correlated with tumor progression. Therefore, there is an urgent need for segmentation and differentiation of distinct tissues to facilitate further clinical research. Lung cancer is the leading cause of cancer-related deaths worldwide. In this challenge, we aim to perform tissue semantic segmentation on Hematoxylin and Eosin (H&E) stained Whole Slide Images (WSIs) of lung adenocarcinoma. The current challenge is that acquiring pixel-level annotations for tissue semantic segmentation is extremely difficult and time-consuming. Inspired by Weakly Supervised Semantic Segmentation (WSS) in computer vision, we decided to use only image-level annotations to perform tissue semantic segmentation. For this challenge, we scanned 67 H&E stained slides from Guangdong Provincial People's Hospital (GDPH) and collected 20 WSIs from The Cancer Genome Atlas (TCGA). Only one WSI was extracted per patient. The goal of this challenge is to achieve pixel-level predictions for three common and clinically meaningful tissue types, namely tumor epithelial tissue, tumor-associated stromal tissue, and normal tissue, using only image-level annotations. Participants will only be provided with image-level annotations (3-class multi-label tags) for training their machine learning algorithms, as well as pixel-level ground truth for validation and testing.

搜集汇总

背景与挑战

背景概述

WSSS4LUAD是一个专注于肺腺癌组织病理学全幻灯片图像（WSI）语义分割的数据集，其关键特点是采用弱监督学习方法，仅提供图像级注释来训练模型进行像素级组织分割，旨在降低标注成本并推动临床研究。数据集包含87个WSI（67个来自广东省人民医院，20个来自TCGA），目标是对肿瘤上皮组织、肿瘤相关基质组织和正常组织进行分割，适用于医学图像分析领域。