April 2011
Volume 52, Issue 14
Free
ARVO Annual Meeting Abstract  |   April 2011
Automated Classification of Pigment Epithelial Detachments using Optical Coherence Tomography
Author Affiliations & Notes
  • Sun Young Lee
    Ophthalmology, Doheny Eye Institute, University of Southern Calif, Los Angeles, California
  • Humberto Ruiz-Garcia
    Ophthalmology, Doheny Eye Institute, University of Southern Calif, Los Angeles, California
  • Paul F. Stetson
    Carl Zeiss Meditec, Dublin, California
  • Florian M. Heussen
    Ophthalmology, Doheny Eye Institute, University of Southern Calif, Los Angeles, California
  • Yanling Ouyang
    Ophthalmology, Doheny Eye Institute, University of Southern Calif, Los Angeles, California
  • SriniVas R. Sadda
    Ophthalmology, Doheny Eye Institute, University of Southern Calif, Los Angeles, California
  • Footnotes
    Commercial Relationships  Sun Young Lee, None; Humberto Ruiz-Garcia, None; Paul F. Stetson, Carl Zeiss Meditec (E); Florian M. Heussen, None; Yanling Ouyang, None; SriniVas R. Sadda, Carl Zeiss Meditec (F), Heidelberg Engineering (C), Optovue, Inc. (F), Topcon Medical Systems (P, R)
  • Footnotes
    Support  None
Investigative Ophthalmology & Visual Science April 2011, Vol.52, 1317. doi:
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Sun Young Lee, Humberto Ruiz-Garcia, Paul F. Stetson, Florian M. Heussen, Yanling Ouyang, SriniVas R. Sadda; Automated Classification of Pigment Epithelial Detachments using Optical Coherence Tomography. Invest. Ophthalmol. Vis. Sci. 2011;52(14):1317.

      Download citation file:


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

      ×
  • Supplements
Abstract

Purpose: : To assess the accuracy of classification of pigment epithelial detachments (PED) using an automated algorithm applied to spectral-domain optical coherence tomography (SD-OCT) scans.

Methods: : Cirrus SD-OCT volume scans (512 x 128) were retrospectively collected from 35 eyes of 24 patients with PEDs in the setting of age-related macular degeneration (AMD) or central serous chorioretinopathy (CSCR). PEDs were automatically detected using the Cirrus OCT RPE Analysis, and among these, a total of 186 PEDs from the 35 eyes were randomly selected for subsequent manual classification by an OCT grader into one of three categories: serous, drusenoid or fibrovascular. The selected PEDs were then analyzed using an automated algorithm which considered the mean internal reflectivity/intensity of all A-scans within the PED as well as the standard deviation (SD, a measure of homogeneity). A-scans in the PED were automatically segregated into one of three categories: low mean intensity and low SD (Type 1 or serous-like), high mean intensity and low SD (Type 2 or drusenoid-like), and high mean intensity and high SD (Type 3 or fibrovascular-like). Based on the predominant A-scan type within the PED, PEDs were automatically classified into the same three categories used by the human grader. Automated classification was repeated after normalizing the A-scans for intensity across the entire volume scan. Automated results were compared to the human grader assessment which was considered to be the gold standard.

Results: : Among the 186 PEDs, the OCT grader classified 8 as serous, 131 as drusenoid, and 47 as fibrovascular. The automated algorithm classified 7 as serous (87% sensitivity and 99% specificity), 108 as drusenoid (79% sensitivity and 93% specificity), and 71 as fibrovascular (91% sensitivity and 80% specificity). The lower sensitivity for detection of cases with drusenoid PEDs, with mis-classification of these cases as fibrovascular PEDs, appeared to be due to RPE hyperplasia and migration into the retina which created areas of apparent in homogeneity and hyporeflectivity within the PED.

Conclusions: : Automated analysis of PEDs on OCT appeared to be sensitive for detecting serous and fibrovascular PEDs but less so for drusenoid PEDs. Refinement of algorithms to account for the effects of intraretinal RPE migration may improve the sensitivity and specificity of these approaches. Automated classification and quantification of PEDs may be a useful tool in future studies of AMD and related diseases.

Keywords: imaging/image analysis: clinical • retinal pigment epithelium 
×
×

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.

×