September 2016
Volume 57, Issue 12
Open Access
ARVO Annual Meeting Abstract  |   September 2016
Automatic Retinal Surface Segmentation in 3D OCT scans with Geographic Atrophy
Author Affiliations & Notes
  • Junjie Bai
    Electrical & Computer Engineering, The University of Iowa, Iowa City, Iowa, United States
  • Zhihong Hu
    Doheny Image Analysis Laboratory, Doheny Eye Institute, Los Angeles, California, United States
  • Yue Shi
    Doheny Image Reading Center, Doheny Eye Institute, Los Angeles, California, United States
  • Srinivas R Sadda
    Ophthalmology, Doheny Eye Institute - UCLA, Los Angeles, California, United States
  • Xiaodong Wu
    Electrical & Computer Engineering, The University of Iowa, Iowa City, Iowa, United States
    Radiation Oncology, The University of Iowa, Iowa City, Iowa, United States
  • Footnotes
    Commercial Relationships   Junjie Bai, None; Zhihong Hu, None; Yue Shi, None; Srinivas Sadda, Allergan (F), Allergan (C), Avalanche (C), Bayer (C), Carl Zeiss Meditec (F), Genetech (F), Genetech (C), Iconic (C), Novartis (C), Optos (F), Optos (C), Regeneron (C), Stem Cells Inc (C), Thrombogenics (C); Xiaodong Wu, None
  • Footnotes
    Support  None
Investigative Ophthalmology & Visual Science September 2016, Vol.57, 5945. doi:
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Junjie Bai, Zhihong Hu, Yue Shi, Srinivas R Sadda, Xiaodong Wu; Automatic Retinal Surface Segmentation in 3D OCT scans with Geographic Atrophy. Invest. Ophthalmol. Vis. Sci. 2016;57(12):5945.

      Download citation file:


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

      ×
  • Supplements
Abstract

Purpose : Geographic atrophy (GA), with loss of the retinal pigment epithelium (RPE) and photoreceptors, is a manifestation of the advanced or late-stage of age-related macular degeneration (AMD), and is increasingly the main cause of vision loss in patients. Current clinical practice typically only uses 2D fundus autofluorescence (FAF) image to quantify the area of GA. The purpose of this study is to develop a fully automatic retinal layer segmentation method in optical coherence tomography (OCT) scans, which can be used to directly assess the GA severity in 3D volume.

Methods : FAF and OCT scans were taken for 12 subjects with late-stage AMD and evidence of GA. Each FAF image contains 768 x 768 pixels with overall physical size of 8.85mm x 8.85mm. Each OCT scan consists of 1024x512x128 (depth x A-scans x B-scans) voxels, with overall physical imaging size of 2mm x 6mm x 6 mm. For each subject, one eye was randomly chosen for subsequent analysis. Manual tracings are obtained for three retinal surfaces in OCT scans: inner-outer segment (IS-OS), Inner RPE and Bruch’s membrane (BM), within region affected by GA.

A fully automated segmentation method is proposed to segment the three surfaces in 3D OCT scans incorporating corresponding FAF image information. First a partial choroid layer OCT projection image is obtained by a double-surface graph search scheme (Hu et al., IOVS 2013). This projection image is then used to register FAF and OCT scans to align each FAF image pixel to an A-scan in OCT. A multi-scale graph search method is performed to simultaneously segment the three target surfaces, using an adaptive surface distance constraint inferred from the registered FAF images.

Results : Fig. 1 illustrates the segmentation results of the proposed method. The segmentation aligns well with the manual tracings. The error between the automatic segmentation and the manual tracing is quantitatively measured by the vertical difference along each A-scan. Table. 1 shows that the unsigned surface position error is within 10μm.

Conclusions : A fully automated retinal surface segmentation method for 3D volumetric OCT scans is proposed, which enables GA severity assessment directly in 3D instead of the current 2D clinical routine.

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

 

Fig. 1. Segmentation of 3 surfaces (blue IS-OS, green Inner RPE and red BM)

Fig. 1. Segmentation of 3 surfaces (blue IS-OS, green Inner RPE and red BM)

 

Table. 1. Surface position error along A-scan

Table. 1. Surface position error along A-scan

×
×

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.

×