Lung Segmentation Data Kit

Lung Segmentation Data Kit is a collection of 385 Chest X-Rays (CXRs) and corresponding lung masks included in both Japanese Society of Radiological Technology (JSRT) and Montgomery dataset. The CXRs included in Montgomery set, as well as their masks, are augmented and resized to the standard JSRT resolution of 2048x2048 pixels. The intensity values of CXR images included in JSRT set are normalised to the intensity values of images included in Montgomery set. The dataset was created in order to train and test Machine Learning based Lung Segmentation Methods on a single unified and denoted dataset instead of two separate ones. The Data Kit provides the end user with more diverse but unified dataset that resembles to the real-life situations.

肺部分割数据集套件（Lung Segmentation Data Kit）整合了来自日本放射技术学会（Japanese Society of Radiological Technology, JSRT）数据集与蒙哥马利（Montgomery）数据集的共385张胸部X光片（Chest X-Rays, CXRs）及其对应的肺部掩码（lung masks）。其中，蒙哥马利数据集中的胸部X光片及其对应掩码均经过数据增强处理，并被统一调整至JSRT数据集的标准分辨率2048×2048像素；JSRT数据集中的胸部X光片像素强度值则被归一化至蒙哥马利数据集图像的强度分布。本数据集的构建初衷，是为研究人员提供单一统一且带标注的标准化数据集，用于训练与测试基于机器学习（Machine Learning）的肺部分割方法，无需分别使用两个独立数据集。该数据集套件可为终端用户提供更具多样性且贴合真实临床场景的统一化数据集。

搜集汇总

背景与挑战

背景概述

Lung Segmentation Data Kit是一个包含385张胸部X光片及其肺部掩模的数据集，整合了JSRT和Montgomery数据集，并进行了标准化处理。该数据集旨在为机器学习肺部分割方法提供统一的训练和测试数据，模拟真实场景。