LungHist700: A Dataset of Histological Images for Deep Learning in Pulmonary Pathology

We present a comprehensive dataset of 691 histopathological images of the lungs, encompassing three categories: adenocarcinomas, squamous cell carcinomas, and normal lung tissues from 45 patients. These images are of high resolution (1200x1600 pixels) and are subdivided into three levels of differentiation for both pathological sets: well differentiated, moderately differentiated, and poorly differentiated. This provides a total of seven classes for classification purposes. A significant portion of the images are at 20x magnification, while others are at 40x magnification, reflecting the diversity of histopathological samples encountered in real clinical practice.本研究构建了一个包含691张肺部组织病理学图像的综合数据集，涵盖三类核心样本：腺癌（adenocarcinoma）、鳞状细胞癌（squamous cell carcinoma）以及来自45名患者的正常肺组织。所有图像均为1200×1600像素的高分辨率格式，且两类肿瘤样本均按照细胞分化程度划分为三个等级：高分化、中分化与低分化。结合上述分层后的肿瘤亚类与正常肺组织类别，该数据集总计包含7个可供分类任务使用的样本类别。此外，数据集内大部分图像的放大倍数为20倍，其余为40倍，能够充分反映真实临床场景中常见的组织病理学样本的多样性。

搜集汇总

背景与挑战

背景概述

LungHist700是一个包含691张高分辨率肺部病理图像的深度学习数据集，涵盖三种组织类型(腺癌、鳞癌、正常)和三种分化程度，共7个分类类别，具有20x和40x两种临床相关放大倍率。该数据集特别适用于肺部病理学的机器学习研究。