MS-CXR

Ms-cxr数据集提供了1162图像-句子对的边界框和相应的短语，收集了八个不同的心肺放射学发现，每个发现的配对数量大致相等。该数据集是对现有MIMIC-CXR v.2数据集的补充，包括: 1.审查和编辑边界框和短语 (1026对边界框/句子); 和2.手动边界框标签从头开始 (136对边界框/句子)。

The Ms-CXR dataset provides bounding boxes and corresponding phrases for 1162 image-sentence pairs, and encompasses eight distinct cardiopulmonary radiological findings with roughly equal numbers of pairs for each finding. This dataset serves as a complement to the existing MIMIC-CXR v.2 dataset, including two components: 1. Reviewed and edited bounding boxes and phrases (1026 bounding box/sentence pairs); and 2. Manual bounding box annotations created from scratch (136 bounding box/sentence pairs).

搜集汇总

背景与挑战

背景概述

MS-CXR数据集是一个专注于心肺放射学发现的图像-句子对数据集，包含1162个配对，覆盖八个不同发现类型，每个类型数量均衡。它提供了边界框和短语标注，部分数据来自对现有MIMIC-CXR数据集的审查编辑，部分为手动新标注，旨在支持医学影像分析研究。