June 2015
Volume 56, Issue 7
ARVO Annual Meeting Abstract  |   June 2015
Reliability of a Manual Procedure for Marking the EZ Band Endpoint Location
Author Affiliations & Notes
  • Rithambara Ramachandran
    Psychology, Columbia University, New York, NY
  • Cindy Cai
    Psychology, Columbia University, New York, NY
  • Dongwon Lee
    Psychology, Columbia University, New York, NY
  • Benjamin Epstein
    Psychology, Columbia University, New York, NY
  • Kirsten G Locke
    Retina Foundation of the Southwest, Dallas, TX
  • David G Birch
    Retina Foundation of the Southwest, Dallas, TX
  • Donald Hood
    Psychology, Columbia University, New York, NY
    Ophthalmology, Columbia University, New York, NY
  • Footnotes
    Commercial Relationships Rithambara Ramachandran, None; Cindy Cai, None; Dongwon Lee, None; Benjamin Epstein, None; Kirsten Locke, None; David Birch, Acucela (C), Neurotech, USA (C), QLT (C); Donald Hood, Topcon, In (F)
  • Footnotes
    Support None
Investigative Ophthalmology & Visual Science June 2015, Vol.56, 5280. doi:
  • Views
  • Share
  • Tools
    • Alerts
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Rithambara Ramachandran, Cindy Cai, Dongwon Lee, Benjamin Epstein, Kirsten G Locke, David G Birch, Donald Hood; Reliability of a Manual Procedure for Marking the EZ Band Endpoint Location. Invest. Ophthalmol. Vis. Sci. 2015;56(7 ):5280.

      Download citation file:

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

  • Supplements

Purpose: The ends of the ellipsoid zone (EZ) band, aka IS/OS border, on frequency domain optical coherence tomography (fdOCT) holds promise as an effective measure for tracking disease progression in clinical studies of retinitis pigmentosa (RP).[1-4] Here we assess the intra- and inter-grader agreement for marking the EZ band endpoint location (EPL).

Methods: After a training session, 3 graders (1 experienced, 2 inexperienced) marked the EPLs on a random sample of horizontal line scans (Spectralis) from 45 RP patients. The average mean deviation of their 30-2 visual field was -22.5 dB (range -2.2 to -33.6dB), and their log MAR ranged from 0 to 0.8. The graders marked the EPLs on the same scans 1 month later. To assess intra- and inter-grader variability, intra-class correlation coefficients (ICC) and signed and absolute differences between the EPL marked by one grader and the average EPL marked by the other two graders [MEAN(ΔEPL)] were calculated. These results were compared to the EPLs marked by an automated segmentation algorithm.[5]

Results: Intra-grader agreement was very good (ICC, 0.99) and the marked EPLs were not statistically significant different (paired t-test) between the two times for any of the graders. The mean signed difference between the two times was 0.07° ± 0.54° and the mean absolute difference was 0.28° ± 0.46°. Likewise, the EPLs did not differ significantly across graders and the mean ICC values were 0.98 (time 1) and 0.97 (time 2). The average signed MEAN(ΔEPL) was 0.01° ± 0.57° and averaged absolute MEAN(ΔEPL) was 0.35° ± 0.45°. The automated algorithm correctly identified the EZ band in only 33 of 45 scans. In 12 scans, the algorithm segmented the proximal retinal pigment epithelium border instead. For the 33 correctly segmented scans, the mean signed and absolute difference between the automated and marked EPL was -0.03° ± 1.40° and 0.93° ± 1.04° respectively.

Conclusions: With training, the inter- and intra-grader reproducibility of manually marked EPLs of the EZ band was excellent. The EPL confidence intervals, about 2°, are smaller than the 6° spacing of the typical visual field used. For clinical studies, the currently available automated segmentation algorithms, if used, will need to be corrected manually.<br /> 1. Hood et al., BOE, 2011; 2. Birch et al., JAMA, 2013; 3. Ramachandran et al., TVST, 2013; 4. Cai et al., IOVS, 2014; 5. Yang et al., BOE, 2011.


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.