May 2006
Volume 47, Issue 13
Free
ARVO Annual Meeting Abstract  |   May 2006
Digital Filters for Correcting Common Sources of Error in Automated Macular Image Analysis
Author Affiliations & Notes
  • T. Smith
    Ophthalmology, Columbia University, New York, NY
  • K. Uy
    Ophthalmology, Columbia University, New York, NY
  • M. Busuioc
    Ophthalmology, Columbia University, New York, NY
  • D. Despriet
    Ophthalmology, Erasmus College, Rotterdam, The Netherlands
  • C.C. K. Klaver
    Ophthalmology, Erasmus College, Rotterdam, The Netherlands
  • Footnotes
    Commercial Relationships  T. Smith, None; K. Uy, None; M. Busuioc, None; D. Despriet, None; C.C.K. Klaver, None.
  • Footnotes
    Support  NEI R01 EY015520 HIGHWIRE EXLINK_ID="47:5:2161:1" VALUE="EY015520" TYPEGUESS="GEN" /HIGHWIRE –01
Investigative Ophthalmology & Visual Science May 2006, Vol.47, 2161. doi:
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      T. Smith, K. Uy, M. Busuioc, D. Despriet, C.C. K. Klaver; Digital Filters for Correcting Common Sources of Error in Automated Macular Image Analysis . Invest. Ophthalmol. Vis. Sci. 2006;47(13):2161.

      Download citation file:


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

      ×
  • Supplements
Abstract

Purpose: : Digital identification of macular drusen can be confounded both by photographic artifacts and non–drusen sources of intrinsic macular reflectance. Herein we report five new corrective algorithms combined with our published drusen detection algorithm (DDA, [1] Smith et al, Arch Ophthal, 2005;123:200–206).

Methods: : 389 color slides graded by experts from the Netherlands genetic isolate study (Erasmus Rucphen Family) were digitized and analyzed with the DDA for the presence or absence of large drusen. A subset of 60 of 200 images with discrepancies between human gradings and the DDA was chosen for expert adjudication and exhaustive scrutiny for causes of the discrepancies. Corrective filters were then constructed from red, green and blue channel information and morphological criteria. The images were re–processed using the appropriate correction(s).

Results: : From the 60 images with disagreements, 32 were adjudicated for the human and 28 for the DDA. After eliminating borderline drusen size disagreements, DDA errors were all false positives: 14 photographic artifacts (dust spots and lens flare), prominent choroid (8), hypopigmentations (7), nerve fiber layer reflections (7), and confluent intermediate drusen identified as large drusen (3). After filter corrections (photo artifacts: blue channel, choroid: red channel, hypopigmentation: combined red and green discriminants, nerve fiber layer and intermediate drusen clusters: morphologic criteria), two hypopigmentation errors remained.

Conclusions: : Automated grading of macular images for drusen in a clinical setting may be significantly improved by filters correcting common sources of error. Distinguishing hypopigmentation from drusen is difficult. Filter optimization on larger datasets is needed.

Keywords: imaging/image analysis: clinical • drusen • age-related macular degeneration 
×
×

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.

×