Ureteroscopy Lumen Segmentation Dataset

This is a dataset for lumen segmentation. The dataset is composed of 1,754 endoscopic images from 23 patients undergoing Ureteroscopy. The dataset is composed of a total 2,187 endoscopic images with its respective masks highlighting the lumen in .png format. The dataset is divided by folders in train/val/test, train and val folders are divided in two sub-folders (image / label) For each image there is a corresponding mask with the same name. The training and validation data is the one described in [1]. The test folder is composed of 3 patient cases, test_01, test_02 and test_03. For more details please check the referred publication associated with this dataset. 

J. F. Lazo et al., "A Lumen Segmentation Method in Ureteroscopy Images based on a Deep Residual U-Net architecture," 2020 25th International Conference on Pattern Recognition (ICPR), Milan, Italy, 2021, pp. 9203-9210, doi: 10.1109/ICPR48806.2021.9411924.

本数据集用于管腔分割任务。数据集包含23名接受输尿管镜检查（Ureteroscopy）的患者的1754张内镜图像，而数据集整体共计2187张内镜图像，并附带对应掩码（mask），所有掩码以PNG格式存储，用于高亮标注管腔区域。数据集按文件夹划分为训练集（train）、验证集（val）与测试集（test）；训练集与验证集文件夹均分为图像（image）与标注（label）两个子文件夹，每张图像均配有同名的对应掩码。训练集与验证集的数据与文献[1]中描述的内容一致。测试集文件夹包含3个患者病例，分别为test_01、test_02与test_03。如需了解更多细节，请查阅本数据集关联的参考文献。

J. F. Lazo 等. 基于深度残差U-Net架构的输尿管镜图像管腔分割方法[C]//2020第25届国际模式识别会议（ICPR）论文集. 意大利米兰: 2021: 9203-9210. DOI: 10.1109/ICPR48806.2021.9411924.