September 2016
Volume 57, Issue 12
Open Access
ARVO Annual Meeting Abstract  |   September 2016
Automated Diagnosis and Segmentation of Branch Retinal Artery Occlusion in SD-OCT
Author Affiliations & Notes
  • Jingyun Guo
    Soochow University, Suzhou, China
  • Xinjian Chen
    Soochow University, Suzhou, China
  • Footnotes
    Commercial Relationships   Jingyun Guo, None; Xinjian Chen, None
  • Footnotes
    Support  Grant 2014CB748600
Investigative Ophthalmology & Visual Science September 2016, Vol.57, 5350. doi:
  • Views
  • Share
  • Tools
    • Alerts
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Jingyun Guo, Xinjian Chen; Automated Diagnosis and Segmentation of Branch Retinal Artery Occlusion in SD-OCT. Invest. Ophthalmol. Vis. Sci. 2016;57(12):5350.

      Download citation file:

      © ARVO (1962-2015); The Authors (2016-present)

  • Supplements

Purpose : Branch retinal artery occlusion (BRAO) is caused by closure of the vessel, disrupting nutrition supply of the corresponding retinal area and resulting in edema because of ischemia and anoxia. Quantitative assessment of the volume of blocked regions on is needed to evaluate the severity.

Methods : 23 subjects diagnosed with branch retinal artery occlusion (12 of acute phase and 11 of chronic phase) were included and underwent macular-centered SD-OCT (Topcon, 512×64×480 voxels, 11.72×93.75×3.50µm3, or 512×128×480 voxels, 11.72×46.88×3.50µm3). Automated diagnosis and segmentation of BRAO was realized as follows. First, an AdaBoost classifier is designed to differentiate BRAO of acute phase, BRAO of chronic phase and normal retina. Then, BRAO regions of acute phase and chronic phase are segmented separately. To segment BRAO in chronic phase, a thickness model is built. While for segmenting BRAO in acute phase, a two-step segmentation is performed: initialization and segmentation. During the initialization phase, the Bayesian posterior probability is utilized to compute the probability map. In the segmentation phase, an advanced GSGC method [1] is applied, in which initialization results are fully utilized as the constraints when constructing the energy function. BRAO regions were labeled in each slice of all subjects as the ground truth, under the supervision of an experienced ophthalmologist. The segmentation results were compared with the ground truth.

Results : Patients with BRAO can be diagnosed and the blocked regions can be segmented automatically using the proposed method (Fig. 1). The accuracy of AdaBoost classifier is 94.3%, the true positive volume fraction (TPVF) and the false positive volume fraction (FPVF) for acute phase were 91.1%, 5.5%; for chronic phase were 90.5%, 8.7%, respectively.

Conclusions : Automated diagnosis and segmentation of BRAO has been achieved, which provides shape, position, and size information to ophthalmologist. We are starting to evaluate the correlation between BRAO volume and visual ability.

This is an abstract that was submitted for the 2016 ARVO Annual Meeting, held in Seattle, Wash., May 1-5, 2016.




This PDF is available to Subscribers Only

Sign in or purchase a subscription to access this content. ×

You must be signed into an individual account to use this feature.