JSRT

该数据集名为JSRT，包含了五个可能存在像素级重叠的类别（右侧/左侧锁骨、右侧/左侧肺、心脏），并用于多标签解剖分割。此外，该数据集还扩展了细粒度注释，共计72个标签，对应于如心脏、肺和肋骨等超类别。规模方面，包含了两套图像集（123/124幅图像）。该数据集的任务是进行多标签解剖分割。

This dataset is named JSRT. It encompasses five categories with potential pixel-level overlaps: right/left clavicles, right/left lungs, and the heart, and is designed for multi-label anatomical segmentation. Furthermore, this dataset is extended with fine-grained annotations, totaling 72 labels corresponding to super-categories such as the heart, lungs, and ribs. In terms of scale, it includes two image sets with 123 and 124 images respectively. The task of this dataset is multi-label anatomical segmentation.

搜集汇总

背景与挑战

背景概述

JSRT数据集是一个包含154张有肺结节和93张无肺结节的胸部X光片数字图像数据库，图像分辨率为2048x2048，12位灰度。数据集还包括患者年龄、性别、结节位置等附加信息，并提供一个迷你版本供初学者学习深度学习技术。