July 2018
Volume 59, Issue 9
Open Access
ARVO Annual Meeting Abstract  |   July 2018
Automated Retinal Layer Segmentation Algorithm for OCT Images: A Validation Study
Author Affiliations & Notes
  • Priyanka Roy
    School of Optometry & Vision Science, University of Waterloo, Waterloo, Ontario, Canada
    Department of Systems Design Engineering, University of Waterloo, Waterloo, Ontario, Canada
  • Mohana Kuppuswamy Parthasarathy
    School of Optometry & Vision Science, University of Waterloo, Waterloo, Ontario, Canada
  • John S. Zelek
    Department of Systems Design Engineering, University of Waterloo, Waterloo, Ontario, Canada
  • Vasudevan Lakshminarayanan
    School of Optometry & Vision Science, University of Waterloo, Waterloo, Ontario, Canada
    Department of Systems Design Engineering, University of Waterloo, Waterloo, Ontario, Canada
  • Footnotes
    Commercial Relationships   Priyanka Roy, None; Mohana Kuppuswamy Parthasarathy, None; John Zelek, None; Vasudevan Lakshminarayanan, None
  • Footnotes
    Support  NSERC Discovery Grant
Investigative Ophthalmology & Visual Science July 2018, Vol.59, 1678. doi:
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Priyanka Roy, Mohana Kuppuswamy Parthasarathy, John S. Zelek, Vasudevan Lakshminarayanan; Automated Retinal Layer Segmentation Algorithm for OCT Images: A Validation Study. Invest. Ophthalmol. Vis. Sci. 2018;59(9):1678.

      Download citation file:


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

      ×
  • Supplements
Abstract

Purpose : Segmentation of optical coherence tomography (OCT) images is crucial for investigation of individual layers of the retina for detection of possible pathologies. However, the task of segmentation is confounded by factors such as poor image contrast and speckle noise. Also, the time and subjectivity involved in manual segmentation (MS) limits the applicability of OCT imaging. It is therefore essential to develop a robust automated segmentation (AS) algorithm. A validation study was performed to testify the segmentation results from a novel automated graph-based algorithm by comparing retinal layer thicknesses from AS with that from MS.

Methods : A shortest path-based graph-search technique was implemented on a de-identified dataset of 10 macular OCT (M-OCT) images and 10 OCT images around the optic disc (OD-OCT) from 20 healthy subjects, to segment 7 retinal boundaries and determine the thickness of the 6 layers thus segmented. The AS was preceded by use of simple Gaussian filters to de-noise the OCT images. The MS was done by an expert clinician on a specially designed graphical user interface. The layer thickness values obtained from MS were used to validate the efficacy of AS by assessing the mean difference between the thickness values given by AS and MS for each layer over the entire dataset.

Results : The unsigned mean differences (UMD) between the layer thickness values (in µm) from AS and MS for the total retina were 2.13±0.94 and 1.36±0.19 for M-OCT and OD-OCT images respectively. The layer-wise UMD for M-OCT images were 2.04±0.76, 1.15±0.79, 2.22±0.31, 2.19±0.27, 1.24±0.40, 2.21±0.43, and that for OD-OCT images were 1.27±0.38, 1.66±0.59, 2.24±0.42, 1.31±0.04, 2.33±0.12, 1.23±0.93, for layers 1 through 6, respectively. The average segmentation time for each image was 2.4s for AS and 357.3s for MS (64-bit Win10, core i5, 8GB RAM).

Conclusions : This high-speed graph-based AS algorithm accurately segmented and computed the thicknesses of 6 retinal layers, delineating 7 boundaries in SD-OCT images. Application of the AS algorithm would considerably reduce time and overhead cost associated with MS of OCT images. Furthermore, from the retinal layer thickness values computed by AS in both M-OCT and OD-OCT, clinicians would be able to detect the presence of possible pathologies, involving macula (such as macular degeneration) and optic nerve head (such as glaucoma), that typically alter retinal layer thicknesses.

This is an abstract that was submitted for the 2018 ARVO Annual Meeting, held in Honolulu, Hawaii, April 29 - May 3, 2018.

 

×
×

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.

×