April 2014
Volume 55, Issue 13
Free
ARVO Annual Meeting Abstract  |   April 2014
Active contour detection for the segmentation of optical coherence tomography images of the retina
Author Affiliations & Notes
  • Gabor Mark Somfai
    Department of Ophthalmology, Semmelweis University, Budapest, Hungary
  • Molnar Jozsef
    MTA SZTAKI, Budapest, Hungary
  • Dmitry Chetverikov
    MTA SZTAKI, Budapest, Hungary
  • Delia DeBuc
    Bascom Palmer Eye Institute, University of Miami, Miller School of Medicine, Miami, FL
  • Footnotes
    Commercial Relationships Gabor Somfai, None; Molnar Jozsef, None; Dmitry Chetverikov, None; Delia DeBuc, None
  • Footnotes
    Support None
Investigative Ophthalmology & Visual Science April 2014, Vol.55, 4793. doi:
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Gabor Mark Somfai, Molnar Jozsef, Dmitry Chetverikov, Delia DeBuc; Active contour detection for the segmentation of optical coherence tomography images of the retina. Invest. Ophthalmol. Vis. Sci. 2014;55(13):4793.

      Download citation file:


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

      ×
  • Supplements
Abstract

Purpose: Recently, active contours have been successfully applied to retinal optical coherence tomography (OCT) image segmentation. Our aim was to describe a novel segmentation algorithm based on active contour detection and test it on rodent OCT images.

Methods: A fast algorithm was developed based on row projections in a sliding window, providing initial borders (“Comb” method). These borders were then used by a slower but more precise variational algorithm that iteratively refines the borders (“Complex” method). Mouse eye retina images acquired by a custom-built spectral-domain OCT device were manually segmented by two independent experts, the test data set containing 20 images. The vitreo-retinal border (VRB), the inner and outer borders for the retinal pigment epithelium (RPEi and RPEo, respectively) were segmented. Then, first the Comb method was used for the segmentation of the same dataset, followed by the Complex method, including the variational refinement. The results of the two methods were compared to the manual segmentation by the mean absolute deviation in pixels from the manual segmentation by unpaired t-test.

Results: The highest number of errors was observed for the segmentation of the RPEo. The Complex method provided significantly less errors in the segmentation compared to the Comb method alone (3.75 vs 2.23, 2.41 vs 1.86 and 3.89 vs 2.86 mean deviation for pixels Comb vs. Complex for the VRB, RPEi and RPEo, respectively, p<0.05 for all comparisons).

Conclusions: There was a small difference observed between the manual and automated segmentation methods used in this study. Active contour detection further enhanced by a variational algorithm seems to be a promising segmentation tool that may be used both in animal and human OCT image segmentation.

Keywords: 688 retina • 549 image processing • 552 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.

×