April 2014
Volume 55, Issue 13
Free
ARVO Annual Meeting Abstract  |   April 2014
Automatic choroidal thickness segmentation in optical coherence tomography
Author Affiliations & Notes
  • David Alonso-Caneiro
    Contact Lens and Visual Optics Lab, Queensland University of Technology, Brisbane, QLD, Australia
  • Scott A Read
    Contact Lens and Visual Optics Lab, Queensland University of Technology, Brisbane, QLD, Australia
  • Michael J Collins
    Contact Lens and Visual Optics Lab, Queensland University of Technology, Brisbane, QLD, Australia
  • Footnotes
    Commercial Relationships David Alonso-Caneiro, None; Scott Read, None; Michael Collins, None
  • Footnotes
    Support None
Investigative Ophthalmology & Visual Science April 2014, Vol.55, 4795. doi:
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      David Alonso-Caneiro, Scott A Read, Michael J Collins; Automatic choroidal thickness segmentation in optical coherence tomography. Invest. Ophthalmol. Vis. Sci. 2014;55(13):4795.

      Download citation file:


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

      ×
  • Supplements
Abstract
 
Purpose
 

Choroidal thickness (ChT) provides valuable information regarding the eye’s anatomy and physiology. A fully automatic method to segment the choroid in OCT images is presented and evaluated in a healthy pediatric population (10-16 years of age) with a range of refractive errors.

 
Methods
 

The automated segmentation method uses graph-search theory to delineate the inner choroidal boundary (ICB) and the outer choroidal boundary (OCB) to obtain the ChT profile from OCT images (Fig 1). Before the segmentation, the B-scan is pre-processed to enhance the boundaries and to minimize the artifacts produced by surrounding features. Based on our previous method (Alonso-Caneiro et al. BOE 2013), the technique was tested on a substantially larger dataset containing 3205 B-scans (3x the original dataset), which had been manually segmented by an experienced observer.

 
Results
 

Two common measurements of error were used to evaluate the performance of the automatic method with respect to the manual segmentation, the mean and the [absolute mean] errors. The ICB, which presents a well-defined boundary, had mean measurement errors of 2.33 [2.69]μm, respectively. The less distinct OCB had higher mean measurement errors of 5.51 [13.02]μm. Analysis of the ChT, which is more clinically relevant, revealed errors of -3.18 [12.63]μm. The manual and automatic methods were highly correlated with an r2=0.98 (Fig 2A). The Bland-Altman plot (Fig 2B) illustrates good agreement between the two methods, and does not exhibit an obvious bias or association between the measurement error and the mean ChT. The mean difference between methods was 2.51±17.68μm (95% C.I. 37.88 to -32.85μm). Taking into consideration that the axial resolution is 3.9μm/pixel, the performance of the automatic method shows a close agreement with the experienced human observer.

 
Conclusions
 

The results demonstrate that the proposed method provides robust detection of the ChT in a large dataset of images with a diverse range of choroidal morphologies.

 
 
Fig 1. Four B-scans representing different examples, with typical variations in thickness and contrast. The segmentation based on the manual (yellow) and automatic (red) for the OCBs is shown as well as the automatic segmentation of the ICB (green).
 
Fig 1. Four B-scans representing different examples, with typical variations in thickness and contrast. The segmentation based on the manual (yellow) and automatic (red) for the OCBs is shown as well as the automatic segmentation of the ICB (green).
 
 
Fig 2. Correlation between the automatic and manual segmentation of central ChT (A). Bland-Altman plot of the difference vs. the mean of the two methods for measurements of central ChT (B).
 
Fig 2. Correlation between the automatic and manual segmentation of central ChT (A). Bland-Altman plot of the difference vs. the mean of the two methods for measurements of central ChT (B).
 
Keywords: 549 image processing • 452 choroid • 550 imaging/image analysis: clinical  
×
×

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.

×