The Multi-Label Retinal Diseases (MuReD) dataset consists of 2,451 multi-label retinal fundus images covering 20 categories. These categories include images of Diabetic Retinopathy (DR), NORMAL images, Media Haze (MH), Optic Disc Cupping (ODC), Tessellation (TSLN), Age-Related Macular Degeneration (ARMD), Drusen (DN), Myopia (MYA), Branch Retinal Vein Occlusion (BRVO), Optic Disc Pallor (ODP), Central Retinal Vein Occlusion (CRVO), Choroidal Neovascularization (CNV), Retinitis (RS), Optic Disc Edema (ODE), Laser Scars (LS), Central Serous Retinopathy (CSR), Hypertensive Retinopathy (HTR), Arteriosclerotic Retinopathy (ASR), Chorioretinitis (CR), and Other Diseases (OTHER) images. After thorough curation and cleaning by the authors, 2,208 high-quality labeled images remained, with a default split of 1,764 images for the training set and 444 images for the validation set.
Millions of people worldwide are diagnosed with retinal diseases each year. Early diagnosis of retinal diseases can effectively prevent patients from deteriorating into permanent blindness. In recent years, artificial intelligence technologies, represented by deep learning algorithms, are actively promoting the development of medical imaging. The authors have collected several retinal disease fundus image datasets, and after strict data cleaning, obtained the MuReD dataset. They hope that this dataset can assist researchers in developing models and software for the identification of retinal diseases based on retinal fundus images, thereby aiding doctors in quickly and accurately diagnosing related diseases and analyzing treatment effectiveness.
Dimensions | Modality | Task Type | Anatomical Structures | Anatomical Area | Number of Categories | Data Volume | File Format |
---|---|---|---|---|---|---|---|
2D | Fundus photography | Classification | Retina | Eye | 20 | 2451 | .tif, .png |
Dataset Statistics | size |
---|---|
min | [519,547] |
median | [1389,1392] |
max | [3925,2847] |
Acronym | Full Name | Training | Validation | Total |
---|---|---|---|---|
DR | Diabetic Retinopathy | 396 | 99 | 495 |
NORMAL | Normal Retina | 395 | 98 | 493 |
MH | Media Haze | 135 | 34 | 169 |
ODC | Optic Disc Cupping | 211 | 52 | 263 |
TSLN | Tessellation | 125 | 31 | 156 |
ARMD | Age-Related Macular Degeneration | 126 | 32 | 158 |
DN | Drusen | 130 | 32 | 162 |
MYA | Myopia | 71 | 18 | 89 |
BRVO | Branch Retinal Vein Occlusion | 63 | 16 | 79 |
ODP | Optic Disc Pallor | 50 | 12 | 62 |
CRVO | Central Retinal Vein Occlusion | 44 | 11 | 55 |
CNV | Choroidal Neovascularization | 48 | 12 | 60 |
RS | Retinitis | 47 | 11 | 58 |
ODE | Optic Disc Edema | 46 | 11 | 57 |
LS | Laser Scars | 37 | 9 | 46 |
CSR | Central Serous Retinopathy | 29 | 7 | 36 |
HTR | Hypertensive Retinopathy | 28 | 7 | 35 |
ASR | Arteriosclerotic Retinopathy | 26 | 7 | 33 |
CRS | Chorioretinitis | 24 | 6 | 30 |
OTHER | Other Diseases | 209 | 52 | 261 |
Statistical table of each label and subset of the data set.
Diabetic Retinopathy Data Example.
Normal Data Example.
Media Haze Data Example.
Optic Disc Cupping Data Example.
Tessellation Data Example.
Age-Related Macular Degeneration Data Example.
Drusen Data Example.
The file structure of the dataset is as follows: it includes a training set label file train_data.csv
, a validation set label file val_data.csv
, and a folder named images
that contains all the images.
Multi-Label Retinal Diseases Dataset
├── images
│ ├── (0001)aria_d_26.tif
│ ├── (0001)aria_d_27.tif
│ ├── ...
├── train_data.csv
├── val_data.csv
Manuel Alejandro Rodriguez Rivera (Khalifa University, United Arab Emirates)
Hasan Al-Marzouqi (Khalifa University, United Arab Emirates)
Panos Liatsis (Khalifa University, United Arab Emirates)
Official Website: https://data.mendeley.com/datasets/pc4mb3h8hz/1
Download Link: https://prod-dcd-datasets-cache-zipfiles.s3.eu-west-1.amazonaws.com/pc4mb3h8hz-1.zip
Article Address: https://doi.org/10.48550/arXiv.2207.02335
Publication Date: 2022-07-26
@article{rodriguez2022multi,
title={Multi-label retinal disease classification using transformers},
author={Rodr{\'\i}guez, Manuel Alejandro and AlMarzouqi, Hasan and Liatsis, Panos},
journal={IEEE Journal of Biomedical and Health Informatics},
year={2022},
publisher={IEEE}
}
Original introduction article is here.