[Arxiv]                  [Download]
 

Data Structure

├── Videos
│   └── 1080_crop
├── Sptatial Annotations 
│   └── images
│   └── masks
│   └── metadata
├── Temporal Annotations
│   └── Temporal_Localization
│   └── SVP_Classification
│   └── Peak_and_trough 

Citation

If you find our RVD dataset is useful in your research, please consider cite:

@article{MD2023RVD,
  title={RVD: A Handheld Device-Based Fundus Video Dataset for Retinal Vessel Segmentation},
  author={MD WAHIDUZZAMAN KHAN, Hongwei Sheng, Hu Zhang, Heming Du, Sen Wang, Minas Theodore Coroneo, 
  Farshid Hajati, Sahar Shariflou, Michael Kalloniatis, Jack Phu, Ashish Agar, Zi Huang, Mojtaba Golzan, Xin Yu},
  journal={arXiv preprint arXiv:2307.06577},
  year={2023}
}

Annotations

Spatial Annotations

The spatial annotations help to understand the structure of vessels, the differences between retinal arteries and veins, and the characteristics of each vein and artery.

Binary vessel masks: Binary vessel masks describe the main structures of retinal vessels. They are frequently used in retinal vessel segmentation. They do not distinguish between arteries and veins.

General vein-artery masks: This type of annotation helps to distinguish between retinal arteries and veins, which is essential as many diseases have different effects on arterial and venous structures.

Fine-grained vein-artery masks: We further divide each artery or vein into four segments based on its width, resulting in the final eight-class masks. These masks precisely reflect the granularities of retinal vessels and are highly demanded when detecting ocular diseases.

Temporal Annotations

Spontaneous retinal Venous Pulsations (SVPs) play a crucial role as a biomarker in retina assessments. This type of annotations helps to study SVPs.

SVP Classification

Temporal Localization

Peak and Trough

Existence of SVPs: We provide the annotations to indicate the presence or absence of SVPs in each video, resulting in 335 “SVP-present” videos and 300 “SVP-absent” videos.

Temporal duration of SVPs: We further provide temporal emergence annotations of SVP by indicating the starting and ending frames of retinal vessel fluctuation. The detailed duration of SVP serves two purposes: it acts as a valuable signal to improve the performance of SVP detection tasks and sets a new task for SVP temporal localization.

“Peak” and “Trough” annotations of SVPs: For each ‘‘SVP-present’’ video, masks for frames with the maximal dilation (‘‘peak’’) or maximal contraction (‘‘trough’’) are provided.

Tools

: We provide some codes for data preprocessing, which can be used to generate the split files.

: We also provide the codes to obtain the benchmark results.

Ethics

RVD dataset is collected in accordance with the guidelines of the Tenets of Helsinki. Written consent was obtained from all participants prior to any data collection, and all examination protocols adhered to the tenets of the Declaration of Helsinki.

Our dataset follows the copyright Creative Commons BY-NC-ND 4.0 license.