[Automatic artery/vein classification using a vessel-constraint network for multi-center fundus images](doi: 10.3389/fcell.2021.659941)
Public: Drive, LES, HRF
| Dataset | Number | Resulotion |
|---|---|---|
| DRIVE_AV | 40 | 584 × 565 |
| LES_AV | 22 | 1444×1620, 1958×2196 |
| HRF_AV | 45 | 3304×2336 |
| Tongren_AV | 30 | 1888×2816 |
| Kailuan_AV | 30 | (1588-2112) × (1586-2112) |
The Tongren clinical dataset contains 30 representative retinal fundus images with a 45 ̊ FOV and a resolution of 1888×2816 pixels, within which 20 images were normal, 10 images with moderate cataract or retinal diseases including glaucoma, age related macular degeneration, and retinal vein occlusion. An approval was obtained from the Ethics Committee of Beijing Tongren Hospital. The ocular fundus had been taken with a fundus camera (CR6-45NM Camera, Canon Inc., Ota, Tokyo, Japan). These images were labeled by two experienced ophthalmologists with the ITK-SNAP toolkit (Yushkevich et al., 2006). For each category, a half images are used for training and the rest are used for testing.
The Kailuan database contains 30 images which were collected from participants of the community-based Kailuan Cohort Study (Jiang et al., 2015). These images have different sizes. The minimum, average, and maximum height are 1588, 1902, 2112. The minimum, average, and maximum width are 1586, 1901, 2112. We use 15 images for training and the rest for testing. As well, these images were labeled by experienced ophthalmologist with the ITK-SNAP toolkit (Yushkevich et al., 2006).
If the code is helpful for your research, please consider citing:
doi: 10.3389/fcell.2021.659941
If you wang to use Tongren and Kailuan dataset,please contact ‘hujingfei24@163.com‘