March 2012
Volume 53, Issue 14
Free
ARVO Annual Meeting Abstract  |   March 2012
Model-based Automatic Segmentation Of The Four Outer Retinal Hyper-reflective Bands From Optical Coherence Tomograms (OCT)
Author Affiliations & Notes
  • Kenneth R. Sloan
    Computer and Information Science,
    UAB, Birmingham, Alabama
  • Douglas H. Ross
    Computer and Information Science,
    UAB, Birmingham, Alabama
  • Richard F. Spaide
    VRM Consultants of NY, New York, New York
  • Christine A. Curcio
    Ophthalmology,
    UAB, Birmingham, Alabama
  • Footnotes
    Commercial Relationships  Kenneth R. Sloan, None; Douglas H. Ross, None; Richard F. Spaide, None; Christine A. Curcio, None
  • Footnotes
    Support  NEI EY06109, EyeSight Foundation of Alabama, Research to Prevent Blindness Inc., Macula Foundation
Investigative Ophthalmology & Visual Science March 2012, Vol.53, 4097. doi:
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Kenneth R. Sloan, Douglas H. Ross, Richard F. Spaide, Christine A. Curcio; Model-based Automatic Segmentation Of The Four Outer Retinal Hyper-reflective Bands From Optical Coherence Tomograms (OCT). Invest. Ophthalmol. Vis. Sci. 2012;53(14):4097.

      Download citation file:


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

      ×
  • Supplements
Abstract
 
Purpose:
 

To extract position and thickness from OCT images for the four outer retinal hyper-reflective bands by using a Gaussian model for each band and accounting for overlapping Gaussian responses.

 
Methods:
 

The method uses a Gaussian function to model each of the four outer retinal hyper-reflective bands in an OCT image [1]. The three outermost bands are modeled as the sum of three Gaussian functions to account for the overlap of the individual Gaussian responses [2]. A summed Gaussian model is fit to the data using a standard optimization technique. Initial values for the three Gaussian function parameters (center position, width, and amplitude) are determined using region- and edge-based image processing techniques. Information from high-confidence features in the image are combined with a model of the band structure to estimate parameters in less clear regions. The optimization step confirms and refines these estimates. The innermost hyper-reflective band has an order-of-magnitude lower amplitude. It is segmented in a separate step, taking into account the effects of the adjacent band.

 
Results:
 

This method provides twelve segmentation curves - inner and outer boundaries and a centerline for each band - that collectively define the thickness and position of the four outer retinal hyper-reflective bands. Previously reported automatic methods [3] provide at most five segmentation curves in the same region. This method also provides results where the distance between two anatomical features is so small that they appear to merge in the OCT image.

 
Conclusions:
 

Determining OCT band information based on a Gaussian model that accounts for band overlap provides more information on band position and thickness than previously reported methods. This will allow improved comparison to retinal anatomy[1].1. Spaide & Curcio. Retina 2011. 31:1609.2. Gilgen. Journal of Lightwave Technology 1989. 7:1225.3. Quellec. IEEE Transactions on Medical Imaging 2010. 29:1321.  

 
Keywords: imaging/image analysis: non-clinical • retina • imaging methods (CT, FA, ICG, MRI, OCT, RTA, SLO, ultrasound) 
×
×

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.

×