June 2013
Volume 54, Issue 15
Free
ARVO Annual Meeting Abstract  |   June 2013
Automatic Segmentations of Intra-retinal Layers and Optic Nerve Head in UHR-OCT Images Using Dynamic Programming
Author Affiliations & Notes
  • Shenghai Huang
    Wenzhou Medical College, Wenzhou, China
  • Meixiao Shen
    Wenzhou Medical College, Wenzhou, China
  • xinting liu
    Wenzhou Medical College, Wenzhou, China
  • Lin Leng
    Wenzhou Medical College, Wenzhou, China
  • Fan Lu
    Wenzhou Medical College, Wenzhou, China
  • Footnotes
    Commercial Relationships Shenghai Huang, None; Meixiao Shen, None; xinting liu, None; Lin Leng, None; Fan Lu, None
  • Footnotes
    Support None
Investigative Ophthalmology & Visual Science June 2013, Vol.54, 5540. doi:
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Shenghai Huang, Meixiao Shen, xinting liu, Lin Leng, Fan Lu; Automatic Segmentations of Intra-retinal Layers and Optic Nerve Head in UHR-OCT Images Using Dynamic Programming. Invest. Ophthalmol. Vis. Sci. 2013;54(15):5540.

      Download citation file:


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

      ×
  • Supplements
Abstract
 
Purpose
 

To propose an automated segmentation algorithm for detection of intra-retinal layers and optic nerve head (ONH) on ultra-high resolution optical coherence tomography (UHR-OCT) images.

 
Methods
 

The OCT images were obtained from a custom-built UHR-OCT with 3μm of axial resolution. Ten macular images from 10 eyes of 5 healthy subjects and seven ONH images from 8 eyes of 6 healthy subjects were included for analysis. The layer segmentation algorithm mainly employs the image intensity and gradient information to identify approximate locations of intra-retinal layers and shortest path search based on dynamic programming to refine the boundary locations to yield accurate segmentation results for each individual layer. The cup-to-disc (C/D) ratio of the optic nerve head is determined by the ILM and RPE layers. To verify the accuracy of the algorithm, the boundary positions of automated segmentations were compared with those of manual segmentations. The horizontal C/D ratio determined by the algorithm was also compared with that evaluated by three expert graders.

 
Results
 

The proposed algorithm successfully segmented eight intra-retinal layers on all UHR-OCT macular images and determined the C/D ratio on the UHR-OCT optic nerve head images automatically. Comparison of the retinal layers detected by the automated algorithm and by manual segmentation, the mean differences between the automated and manual detections ranged between 2.1μm and 7.8μm in horizontal meridian and the mean standard deviation was less than 1.4μm. The horizontal C/D ratio determined by the algorithm had good correlations with that determined by three experts (r > 0.80, P < 0.05). And the correlations between the algorithm and manual measurements were comparable with those between the manual measurements evaluated by the three experts (r ranged from 0.82 to 0.91 between manual measurements).

 
Conclusions
 

It was demonstrated that the developed algorithms can detect eight intra-retinal layers with high accuracy. The determined horizontal C/D ratio by the algorithm is comparable with that evaluated by retinal expert. This method provides an objective and quantitative analysis for retina and may hold some potential applications in early diagnosis of retinal diseases and glaucoma.

 
 
Nine Boundaries of the Intra-retinal Layers
 
Nine Boundaries of the Intra-retinal Layers
 
 
Automatic Segmentations of Optic Nerve Head
 
Automatic Segmentations of Optic Nerve Head
 
Keywords: 552 imaging methods (CT, FA, ICG, MRI, OCT, RTA, SLO, ultrasound) • 688 retina • 627 optic disc  
×
×

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.

×