June 2013
Volume 54, Issue 15
Free
ARVO Annual Meeting Abstract  |   June 2013
Automated Choroidal Segmentation in 1µm Swept Source Deep Range Imaging OCT
Author Affiliations & Notes
  • Qi Yang
    Topcon Adv Biomed Imaging Lab, Topcon Medical Systems, Oakland, NJ
  • Charles Reisman
    Topcon Adv Biomed Imaging Lab, Topcon Medical Systems, Oakland, NJ
  • Rithambara Ramachandran
    Psychology, Columbia University, New York, NY
  • Ali Raza
    Psychology, Columbia University, New York, NY
  • Donald Hood
    Psychology, Columbia University, New York, NY
    Ophthalmology, Columbia University, New York, NY
  • Kinpui Chan
    Topcon Adv Biomed Imaging Lab, Topcon Medical Systems, Oakland, NJ
  • Footnotes
    Commercial Relationships Qi Yang, Topcon Medical Systems, Inc. (E); Charles Reisman, Topcon Medical Systems, Inc. (E); Rithambara Ramachandran, None; Ali Raza, None; Donald Hood, Topcon, In (F); Kinpui Chan, None
  • Footnotes
    Support None
Investigative Ophthalmology & Visual Science June 2013, Vol.54, 5500. doi:
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Qi Yang, Charles Reisman, Rithambara Ramachandran, Ali Raza, Donald Hood, Kinpui Chan; Automated Choroidal Segmentation in 1µm Swept Source Deep Range Imaging OCT. Invest. Ophthalmol. Vis. Sci. 2013;54(15):5500.

      Download citation file:


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

      ×
  • Supplements
Abstract
 
Purpose
 

To develop and evaluate a fully automated choroidal-scleral interface (CSI) segmentation algorithm using 1µm wavelength swept source Deep Range Imaging (DRI) OCT.

 
Methods
 

12x9mm2 wide scan volumes of 10 normal subjects and 10 glaucoma patients were acquired using DRI OCT-1 (Topcon Corp., Tokyo, Japan), in which 5 healthy controls and 5 patients each had two repetitions. An algorithm employing a morphological filter and dual-gradient information [1, 2] was used to delineate the inner limiting membrane (ILM), Bruch’s membrane (BM) and the CSI. To test the accuracy of the algorithm, the point-to-point manually segmented border position [3] were subtracted from the automatic segmentation for each B-scan. These difference values (signed) were averaged, as were the absolute (unsigned) values. To test repeatability, the 3D scans of the 10 eyes with two repetitions were automatically segmented, and the intra-class correlation coefficient (ICC), coefficient of variation (CV) and the mean of standard deviation (SD) were calculated.

 
Results
 

For the accuracy analysis, the mean unsigned differences between the manual and automated border positions were: 2.39±6.05 (ILM), 3.40±1.24 (BM), and 20.11±19.42 µm (CSI) and the mean signed differences were: -0.64±5.84 (ILM), 1.53±1.43 (BM), and 4.11±22.98 µm (CSI). For the repeatability analysis, the ICC ranged from 0.91 to 0.84; the CV from 1.6 to 6.5%; and the SD from 3.9 to 15.8µm (Table 1).

 
Conclusions
 

Given that the CSI generally has the lowest contrast among the boundaries segmented in the OCT images, the automated choroidal segmentation demonstrated reasonable repeatability and accuracy on the scans covering both the macular and optic disc regions. The automated segmentation tended to deviate posterior to the manual segmentation in the very thin choroid region, which contributed to the increased mean signed difference between manual and automated segmentation. In the immediate vicinity of optic disc region, the automated choroidal segmentation also had greater error than the other boundaries as the choroid usually thins rapidly as it approaches the optic disc. 1. Yang Q, et al., Opt Express 2010; 2. Yang Q, et al., Biomed Opt Express 2011; 3. Hood D, et al, IOVS, 2011.

 
 
Table 1. Repeatability of 10 eyes (2 repetitions, 5 patients and 5 controls)
 
Table 1. Repeatability of 10 eyes (2 repetitions, 5 patients and 5 controls)
 
Keywords: 452 choroid • 549 image processing  
×
×

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.

×