ROSE(Retinal OCTA SEgmentation dataset)

视网膜 OCT 血管造影血管分割数据集 (ROSE) 现在是开源的。它包括两个子集：ROSE-1 和 ROSE-2。
ROSE-1 子集由来自 39 名受试者（26 名患有疾病，其余为健康对照）的 117 张 OCTA 图像组成，分为 90 张用于训练的图像和 27 张用于测试的图像。每个受试者都有浅表 (SVC)、深部 (DVC) 和视网膜内血管丛的正面血管造影照片，分别包括 SVC 和 DVC (SVC+DVC)。所有 OCTA 扫描均由配备 AngioVue 软件的 RTVue XR Avanti SD-OCT 系统（Optovue，USA）捕获，图像分辨率为 304 × 304 像素。扫描区域为以中央凹为中心的 3×3 mm2 区域。这些血管网络的人工注释由图像专家和临床医生进行分级，然后将他们的共识用作基础。有两种不同类型的血管注释可用：中心线级和像素级注释。

ROSE-2 子集包含由 Heidelberg OCT2 系统和 Spectralis 软件（Heidelberg Engineering，Heidelberg，Germany）获取的 112 只眼睛的 112 张 OCT-A 图像，分为 90 幅图像用于训练和 22 幅图像用于测试。该数据集中的所有图像都是以中央凹为中心的 3×3 mm2 区域内 SVC 的正面血管造影照片，每张图像都被调整为 840×840 像素的灰度图像。所有可见的视网膜血管均由经验丰富的临床眼科医生使用 Matlab (Mathworks R2018, Natwick) 编写的内部程序手动追踪。

The Retinal OCT Angiography Vessel Segmentation Dataset (ROSE) is now open-source. It comprises two subsets: ROSE-1 and ROSE-2.

The ROSE-1 subset consists of 117 OCTA images from 39 subjects (26 with disease, the remaining as healthy controls), split into 90 training images and 27 test images. Each subject has en face angiograms of the superficial vascular plexus (SVC), deep vascular plexus (DVC), and intraretinal vascular plexus, including both SVC and DVC (SVC+DVC). All OCTA scans were acquired using the RTVue XR Avanti SD-OCT system (Optovue, USA) equipped with AngioVue software, with an image resolution of 304 × 304 pixels. The scan area is a 3×3 mm² region centered on the fovea. Manual annotations of these vascular networks were graded by image experts and clinicians, with their consensus used as the ground truth. Two distinct types of vascular annotations are available: centerline-level and pixel-level annotations.

The ROSE-2 subset contains 112 OCT-A images from 112 eyes, acquired using the Heidelberg OCT2 system and Spectralis software (Heidelberg Engineering, Heidelberg, Germany), split into 90 training images and 22 test images. All images in this dataset are en face angiograms of the SVC within a 3×3 mm² region centered on the fovea, and each image has been resized to 840×840 pixels as grayscale images. All visible retinal vessels were manually traced by experienced clinical ophthalmologists using an in-house program written in Matlab (Mathworks R2018, Natwick).

搜集汇总

背景与挑战

背景概述

ROSE是一个用于视网膜OCT血管造影血管分割的开源医疗数据集，包含ROSE-1和ROSE-2两个子集，分别提供不同分辨率和注释类型的OCTA图像，适用于血管分割模型的训练和测试。数据集由多个研究机构于2020年发布，支持中心线级和像素级血管标注，旨在促进视网膜血管分析的研究。