July 2019
Volume 60, Issue 9
Open Access
ARVO Annual Meeting Abstract  |   July 2019
Real-time retinal layer segmentation for high-resolution OCT angiography
Author Affiliations & Notes
  • Worawee Janpongsri
    Simon Fraser University, Burnaby, British Columbia, Canada
  • Morgan Heisler
    Simon Fraser University, Burnaby, British Columbia, Canada
  • Myeong Jin Ju
    Simon Fraser University, Burnaby, British Columbia, Canada
  • Marinko V Sarunic
    Simon Fraser University, Burnaby, British Columbia, Canada
  • Yifan Jian
    Casey Eye Institute, Oregon Health & Science University, Portland, Oregon, United States
  • Footnotes
    Commercial Relationships   Worawee Janpongsri, None; Morgan Heisler, None; Myeong Ju, Seymour Vision Inc. (E); Marinko Sarunic, Seymour Vision Inc. (I); Yifan Jian, Seymour Vision Inc. (I)
  • Footnotes
    Support  None
Investigative Ophthalmology & Visual Science July 2019, Vol.60, 156. doi:
  • Views
  • Share
  • Tools
    • Alerts
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Worawee Janpongsri, Morgan Heisler, Myeong Jin Ju, Marinko V Sarunic, Yifan Jian; Real-time retinal layer segmentation for high-resolution OCT angiography. Invest. Ophthalmol. Vis. Sci. 2019;60(9):156.

      Download citation file:

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

  • Supplements

Purpose : Optical coherence tomography angiography (OCTA) has been widely utilized clinically to examine retinal vasculature and its hierarchical structure. Due to the limited resolution of OCT, the detected vessel size is typically larger, an effect that particularly affects small capillaries. Adaptive Optics (AO) techniques have been applied to OCT to overcome this limitation, and has been demonstrated for high-resolution retinal angiography. Since AO based OCT has a shorter depth-of-focus compared to standard OCT, an effective focus control is crucial to achieve high quality images. We have developed a real-time retinal layer segmentation that provides the effective focus control and direct feedback of aberration correction performance with image-guided AO techniques.

Methods : Our retinal segmentation algorithm is based on a shortest-path graph search applied in a limited region. The Inner Limiting Membrane (ILM) and Inner Segment/Outer Segment (IS/OS) layers are first coarsely segmented, then other retinal layer boundaries (NFL/GCL, INL/OPL, OPL/ONL) are sequentially segmented as shown in Fig.1 (a). The algorithm uses down-sampling of the images to improve the speed for real time applications.

Results : The time for segmentation only takes 0.12 seconds for a single OCT B-scan data (140 x 400). With down-sampling by a factor of 2, the segmentation time reduced to 13 milliseconds which is suitable for integration with real time acquisition. The accuracy of the retinal boundaries segmentation is satisfactory for focus control and/or aberration correction performance with image guided AO technique.
The results of our segmentation algorithm on AO-OCTA en face images is compared with the results using commercial OCTA en face images (PlexElite, Carl Zeiss Meditec) of the same subject are shown in Fig.1 (c) – (f). The AO-OCTA en face provides sharper images of the vasculature with vessel diameters approaching expected results based on histology.

Conclusions : High resolution AO OCT images of the retinal vasculature had a reduction of projection artifacts, which resulted in clear images of the different retinal vessel beds. This provides a better estimation of the vessel index and improves vessel tracing, which could allow improved inspection of vasculopathies.

This abstract was presented at the 2019 ARVO Annual Meeting, held in Vancouver, Canada, April 28 - May 2, 2019.



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.