DRAC 2022
Diabetic Retinopathy Analysis Challenge

News

• [2022.10.11] We released the final ranking of our challenge.
• [2022.10.11] We successfully held the satellite event on September 18th, thanks to all the presenting teams. We put the PPT of the presenting teams here for everyone to learn.
• [2022.09.19] We updated the summay paper publication rules.
  
• [2022.09.15] We have sent an invitation letter to invite some top teams to participate in the satellite event of MICCAI2022.
• [2022.09.13] We updated the paper submission rules.
  
• [2022.09.12] Please note that only the latest submission that will be considered for the final challenge results, so don't forget to submit your best results before the deadline. Also, please read the paper submission rules carefully and prepare your paper as required.
  
• [2022.08.31] In order to ensure that participants have sufficient time for algorithm development and paper writing, we have decided to postpone the deadline for results submission to September 12 and paper submission to October 8.
  
• [2022.08.22] We have updated the ground truths of training set and test set for task 1 (Segmentation). The dataset download link is the same as before. Please make sure you are using the latest ground truth. We also synchronized the ground truth of test set in the grand challenge ranking system. For task 1, Only results submitted after this time are considered valid. You may choose to resubmit your previous results to view the results on the new test set. For your convenience, we have set the limit on the number of submission times for task 1 from 2 to 4 submissions per 2 days.
• [2022.08.22] We have updated the rules and added the part of MICCAI 2022 Challenge Proceedings.
  
• [2022.08.15] Label Correction! In task 3 of DR Grading, the label in training set for 107.png is 0 (the original is 1). We have updated it in the file from the download link.
• [2022.08.08] The submission is now opening, and the submission guidance has been provided, please see the Evaluation.
  
• [2022.08.08] We have released the testing set. The dataset download link is the same as before.
• [2022.08.05] We have updated the Timeline for test set release and result submission, see Timeline.
• [2022.08.02] We have released the training set for Task 1.
• [2022.08.02] We have released the corrected labels in training set for Task 3. The dataset download link is the same as before.

Introduction

The challenge DRAC22 is a first edition associated with MICCAI 2022. Three tasks are proposed (participants can choose to participate in one or all three tasks):

• Task 1: Segmentation of  Diabetic Retinopathy Lesions.
  
• Task 2: Image Quality Assessment.
• Task 3: Diabetic Retinopathy Grading.

Diabetic retinopathy is one of the leading causes of blindness and affects approximately 78% people, with a history of diabetes of 15 years or longer [1]. DR often causes gradual changes in vasculature structure and resulting abnormalities. DR is diagnosed by visually inspecting retinal fundus images for the presence of retinal lesions, such as microaneurysms (MAs), intraretinal microvascular abnormalities (IRMAs), nonperfusion areas and neovascularization. The detection of these lesions is critical to the diagnosis of DR. There have been some works using fundus images for DR diagnosis [2]. With rising popularity, OCT angiography (OCTA) has the capability of visualizing the retinal and choroidal vasculature at a microvascular level in great detail [3]. Specially, swept-source (SS)-OCTA allows additionally the individual assessment of the choroidal vasculature. There are already some works using SS-OCTA to grade for qualitative features of diabetic retinopathy [4-6]. Further, ultra-wide optical coherence tomography angiography imaging (UW-OCTA) modality showed higher burden of pathology in the retinal periphery that was not captured by typical OCTA [7]. Some works already use UW-OCTA on DR analysis [7, 8]. The traditional diagnosis of DR grading mainly relies on fundus photography and FFA, especially for PDR, which seriously endangers vision health. FA is mainly used to detect the presence or absence of new blood vessels. Fundus photography is difficult to detect early or small neovascular lesions. FA is an invasive fundus imaging that cannot be used in patients with allergies, pregnancy, or poor liver and kidney function. The ultra-wide OCTA can non-invasively detect the changes of DR neovascularization, thus it is an important imaging modality to help ophthalmologist diagnose PDR. However, there are currently no works capable of automatic DR analysis using UW-OCTA. In the process of DR analysis, the image quality of UW-OCTA needs to be assessed first, and the images with better imaging quality are selected. Then DR analysis is performed, such as lesion segmentation and PDR detection. Thus, it is crucial to build a flexible and robust model to realize automatic image quality assessment, lesion segmentation and PDR detection. In order to promote the application of machine learning and deep learning algorithms in automatic image quality assessment, lesion segmentation and PDR detection using UW-OCTA images, and promote the application of corresponding technologies in clinical diagnosis of DR, we provide a standardized ultra-wide (swept-source) optical coherence tomography angiography (UW-OCTA) data set for testing the effectiveness of various algorithms. With this dataset, different algorithms can test their performance and make a fair comparison with other algorithms. We believe this dataset is an important milestone in automatic image quality assessment, lesion segmentation and DR grading.

REFERENCES

[1] Tian M , Wolf S , Munk M R , et al. Evaluation of different Swept'Source optical coherence tomography angiography (SSOCTA) slabs for the detection of features of diabetic retinopathy[J]. Acta ophthalmologica, 2019, 98(1).

[2] Dai, L., Wu, L., Li, H. et al. A deep learning system for detecting diabetic retinopathy across the disease spectrum. Nat Commun 12, 3242 (2021). https://doi.org/10.1038/s41467-021-23458-5.

[3] Spaide R F, Fujimoto J G, Waheed N K, et al. Optical coherence tomography angiography[J]. Progress in retinal and eye research, 2018, 64: 1-55.

[4] Schaal K B, Munk M R, Wyssmueller I, et al. Vascular abnormalities in diabetic retinopathy assessed with sweptsource optical coherence tomography angiography widefield imaging[J]. Retina, 2019, 39(1): 79-87.

[5] Stanga P E, Papayannis A, Tsamis E, et al. New findings in diabetic maculopathy and proliferative disease by swept-source optical coherence tomography angiography[J]. OCT Angiography in Retinal and Macular Diseases, 2016, 56: 113-121.

[6] Schaal K B, Munk M R, Wyssmueller I, et al. Vascular abnormalities in diabetic retinopathy assessed with sweptsource optical coherence tomography angiography widefield imaging[J]. Retina, 2019, 39(1): 79-87.

[7] Zhang Q, Rezaei K A, Saraf S S, et al. Ultra-wide optical coherence tomography angiography in diabetic retinopathy[J]. Quantitative imaging in medicine and surgery, 2018, 8(8): 743.

[8] Russell J F, Shi Y, Hinkle J W, et al. Longitudinal wide-field swept-source OCT angiography of neovascularization in proliferative diabetic retinopathy after panretinal photocoagulation[J]. Ophthalmology Retina, 2019, 3(4): 350-361.

How to participate

Participants  should carefully study the challenge rules and follow these steps:

1. Register on the grand-challenge website.
2. Sign in to your account and revisit the challenge webpage.
3. Click on the green "Join" bottom on the top right corner of the webpage.

*Please note that by participating in this challenge you are agreeing to all its rules and policies.