HVDROPDB Datasets for Classification and Segmentation for Research in Retinopathy of Prematurity, Ranjana Agrawal

HVDROPDB_RetCam_Neo_Segmentation and HVDROPDB_RetCam_Neo_Classification are the first datasets to be published for the retinal structure segmentation to identify the Retinopathy of Prematurity (ROP). They are prepared by screening the preterm infants visiting PBMA's H.V. Desai Eye Hospital, Pune with two diverse imaging systems RetCam and Neo. The Segmentation dataset contains sub-datasets for the segmentation of optic disc, blood vessels, and demarcation line/ridge from the fundus images of preterm infants, annotated by a group of ROP experts. Each sub-dataset contains retinal fundus images of premature infants with the ground truths prepared manually to assist researchers in developing an explainable automated ROP screening system. The Classification sub-datasets contain ROP and Normal images.

HVDROPDB_RetCam_Neo_Segmentation与HVDROPDB_RetCam_Neo_Classification是首个面向视网膜结构分割任务、用于识别早产儿视网膜病变（Retinopathy of Prematurity, ROP）的公开数据集。本数据集通过对就诊于浦那PBMA的H.V. Desai眼科医院的早产儿进行筛查采集而来，采用RetCam与Neo两种不同的成像系统获取图像。其中分割数据集包含三个子数据集，分别用于从早产儿视网膜眼底图像中分割视盘、视网膜血管及分界线/脊，由ROP领域专家团队完成标注。各子数据集均配备早产儿视网膜眼底图像与人工制备的金标准标注真值，旨在助力研究人员开发可解释的自动化ROP筛查系统。分类子数据集则包含ROP病变图像与正常对照眼底图像。