Automatic localization of optic disc using modified U-Net

Zaiwang Gu, Shanshan Jiang, Jimmy Lee, Jianyang Xie, Jun Cheng, Jiang Liu

Research output: Chapter in Book/Conference proceedingConference contributionpeer-review

6 Citations (Scopus)


The optic disc (OD) localization plays an important role in the automatic retinal image analysis for many applications such as glaucoma detection, macular localization, and retinal vessel analysis. In this paper, we propose a method based on U-net and Depth-First-Select Graph to accurately and efficiently locate the optic disc. The adopted U-net architecture is based on ResNet-50, and it predicts the center of OD and produces a probability map. Then based on the probability map, we use the Depth-First-Select algorithm to select the brightest and largest region, which is most likely to be the OD. The proposed method is evaluated on the ORIGA and Messidor dataset. Our experiment shows that the proposed method achieves 100% accuracy in ORIGA and 99.83% accuracy in Messidor. It outperforms other optic disc localization algorithms.

Original languageEnglish
Title of host publicationProceedings of 2018 International Conference on Control and Computer Vision, ICCCV 2018
PublisherAssociation for Computing Machinery
Number of pages5
ISBN (Electronic)9781450364706
Publication statusPublished - 15 May 2018
Externally publishedYes
Event2018 International Conference on Control and Computer Vision, ICCCV 2018 - Singapore, Singapore
Duration: 15 Jun 201818 Jun 2018

Publication series

NameACM International Conference Proceeding Series


Conference2018 International Conference on Control and Computer Vision, ICCCV 2018


  • Deep learning
  • Deep-first-select
  • Retinal image analysis
  • U-Net

ASJC Scopus subject areas

  • Software
  • Human-Computer Interaction
  • Computer Vision and Pattern Recognition
  • Computer Networks and Communications


Dive into the research topics of 'Automatic localization of optic disc using modified U-Net'. Together they form a unique fingerprint.

Cite this