search

HzFu / EyeQ

EyeQ : Dataset of Retinal Image Quality Assessment
144 stars 41 forks source link

readme

Eye-Quality (EyeQ) Assessment Dataset

The project web for "Evaluation of Retinal Image Quality Assessment Networks in Different Color-spaces" in MICCAI 2019.

Fundus Enhancement: We also have a related work for "Modeling and Enhancing Low-quality Retinal Fundus Images" in IEEE TMI, 2021. The code is released in Github: Cofe-Net

-Introduction:

Eye-Quality (EyeQ) Assessment Dataset is a re-annotatation subset from the EyePACS dataset for fundus image quality assessment.

EyeQ dataset has 28,792 retinal images with a three-level quality grading (i.e., 'Good', 'Usable' and 'Reject').

Examples of different retinal image quality grades.

Train - - - - - Test - - - - - Total
DR-0 DR-1 DR-2 DR-3 DR-4 All DR-0 DR-1 DR-2 DR-3 DR-4 All
Good 6,342 699 1,100 167 39 8,347 5,966 886 1,354 199 65 8,470 16,817
Usable 1,353 103 283 79 58 1,896 3,201 359 721 145 133 4,559 6,435
Reject 1,544 109 426 87 154 2,320 2,195 153 569 104 199 3,220 5,540
Total 9,239 911 1,809 333 251 12,543 1,1362 1,398 2,644 448 397 16,249 28,792

-Usage:

  1. The original fundus images could be downloaded from EyePACS dataset.
  2. All the original fundus images should be pre-porcessed by 'EyeQ_process_main.py' in folder './EyeQ_preprocess'.
  3. The quality label is in './data' folder, where the 'Label_EyeQ_train.csv' and 'Label_EyeQ_test.csv' are divided by EyePACS, and the 'DR_grade' label is also from EyePACS.
  4. We also release our Multiple Color-space Fusion Network (MCF-Net) based on ResNet121 in './MCFNet' folder.

Note: The trained model of MCF-Net 'DenseNet121_v3_v1.tar' (~112MB) could be download from OneDrive.

-Reference:

If you use this dataset and code, please cite the following papers:

[1] Huazhu Fu, Boyang Wang, Jianbing Shen, Shanshan Cui, Yanwu Xu, Jiang Liu, Ling Shao, "Evaluation of Retinal Image Quality Assessment Networks in Different Color-spaces", in MICCAI, 2019. [PDF] Note: The corrected accuracy score of MCF-Net is 0.8800.

-License:

The code is released under the Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International Public License for NonCommercial use only. Any commercial use should get formal permission first.

Update log: