The ROSE (Retinal OCT-Angiography vessel SEgmentation) dataset is an open-source OCTA retinal blood vessel segmentation dataset, comprising two sub-datasets: ROSE-1 and ROSE-2.
The ROSE-1 dataset includes 117 OCTA images from 39 subjects (26 Alzheimer's patients and 13 healthy subjects). The dataset is divided into 90 images for training and 27 images for testing. Each subject has en face (horizontal cross-sectional) angiograms of the superficial vascular complex (SVC), deep vascular complex (DVC), and the inner retinal vascular plexuses containing both SVC and DVC. The scanning area is a 3×3 square millimeter region centered around the fovea, with an image resolution of 304×304 pixels. ROSE-1 provides two different types of vascular annotations: centerline-level annotations and pixel-level annotations.
The ROSE-2 dataset contains a total of 112 OCT-A images, with 90 for training and 22 for testing. All images in this dataset are en face angiograms of the SVC within a 3×3 square millimeter area centered around the fovea. Only centerline-level annotations are provided.
Retinal vascular changes can assess disease and treatment effectiveness, especially in the foveal region - the part of the retina responsible for sharp central vision. Traditional color fundus photography struggles to capture the subtle vessels around the fovea, while methods like fluorescein angiography, though detailed, carry invasive risks. Optical Coherence Tomography Angiography (OCTA), as a safe technology, provides high-resolution images of the vasculature in the foveal region, essential for the early diagnosis of retinal diseases. Changes revealed by OCTA are also related to neurodegenerative diseases such as Alzheimer's disease, making it a powerful tool for researching and monitoring these conditions.
| Dimensions | Modality | Task Type | Anatomical Structures | Anatomical Area | Number of Categories | Data Volume | File Format |
|---|---|---|---|---|---|---|---|
| 2D | OCTA | Segmentation | Fundus Vessel | Fundus | 1 | 229 | .png .tif |
The image size for ROSE-1 is 304×304.
The image size for ROSE-2 is 512×512.
| Category | Retinal Vessel |
|---|---|
| Number of Images | 117 |
| Availability | 100% |
| Small Vessel Count | 225 |
| Medium Vessel Count | 15928 |
| Large Vessel Count | 22627 |
| Category | Retinal Vessel |
|---|---|
| Number of Images | 112 |
| Availability | 100% |
| Small Vessel Count | 4218 |
| Medium Vessel Count | 19792 |
| Large Vessel Count | 39159 |
Paper Visualization: Annotations of the ROSE-1 Dataset
Paper Visualization: Annotations of the ROSE-2 Dataset
Local Visualization: Visualization of ROSE-1 Images and Annotations.
Local Visualization: Visualization of ROSE-2 Images and Annotations.
The file structure of the ROSE dataset is as follows, divided into two sub-datasets: ROSE-1 and ROSE-2. Each sub-dataset provides train and test folders, with each folder offering img and gt.
data
├── ROSE-1
│ ├── DVC
│ │ ├── test
│ │ │ ├── gt
│ │ │ └── img
│ │ └── train
│ │ ├── gt
│ │ └── img
│ ├── SVC
│ │ ├── test
│ │ │ ├── gt
│ │ │ └── img
│ │ └── train
│ │ ├── gt
│ │ └── img
│ └── SVC_DVC
│ ├── test
│ │ ├── gt
│ │ └── img
│ └── train
│ ├── gt
│ └── img
└── ROSE-2
├── test
│ ├── gt
│ └── original
└── train
├── gt
└── original
Yitian Zhao (Medical and Ophthalmic Imaging Team, Ningbo Institute of Industrial Technology)
Official Website: https://imed.nimte.ac.cn/dataofrose.html
Download Link: https://imed.nimte.ac.cn/dataofrose.html
Article Address: https://arxiv.org/pdf/2007.05201.pdf
Publication Date: 2020
article{ma2021rose:,
title={ROSE: a retinal OCT-angiography vessel segmentation dataset and new model},
author={Ma, Yuhui and Hao, Huaying and Xie, Jianyang and Fu, Huazhu and Zhang, Jiong and Yang, Jianlong and Wang, Zhen and Liu, Jiang and Zheng, Yalin and Zhao, Yitian},
journal={IEEE Transactions on Medical Imaging},
year={2021},
volume={40},
number={3},
pages={928--939},
doi={10.1109/TMI.2020.3042802},
publisher={IEEE}
}
Original introduction article is here.