May 2004
Volume 45, Issue 13
ARVO Annual Meeting Abstract  |   May 2004
Automated drusen quantitaion for clinical trials
Author Affiliations & Notes
  • M.R. Barakat
    Dept Ophthalmology, Scheie Eye Institute, Philadelphia, PA
  • B. Madjarov
    Dept Ophthalmology, Scheie Eye Institute, Philadelphia, PA
  • Footnotes
    Commercial Relationships  M.R. Barakat, None; B. Madjarov, None.
  • Footnotes
    Support  none
Investigative Ophthalmology & Visual Science May 2004, Vol.45, 3017. doi:
  • Views
  • Share
  • Tools
    • Alerts
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      M.R. Barakat, B. Madjarov; Automated drusen quantitaion for clinical trials . Invest. Ophthalmol. Vis. Sci. 2004;45(13):3017.

      Download citation file:

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

  • Supplements

Abstract: : Purpose: Multiple clinical trials currently are quantifying drusen parameters manually in rough scale stratification. This is a labor and time consuming task with limited accuracy. We aim to develop and evaluate a fast and accurate computer–based drusen detection and analysis platform for the purposes of clinical trials. Methods: A Java–enabled, web–based platform for drusen segmentation and measurement was custom developed to facilitate the grading process. Twenty color fundus photographs from patients with early age–related macular degeneration were randomly selected. The slides were digitized with a film scanner at 1000 dpi resolution in three channels. Testing was performed on Pentium II computer with a Windows operating system. The software was evaluated for drusen delineation accuracy as well as for precision in reporting drusen sizes and location, compared to manual grading. In order to improve the robustness, a region of interest was defined by the operator on each image corresponding to the fundus area of predominant drusen location. Adjustment for the variation of the brightness level was incorporated in order to increase the efficiency of drusen detection. Subsequently, an intensity index was defined, based on several scanning cycles of drusen detection to optimize the algorithm’s performance. Results:The software provided a friendly graphical user interface for facile operator interaction. The algorithm displayed a sensitivity of 86% and a positive predictive value of 93% for drusen detection as compared to manual examination of digital images as a standard. The mean number of automated detection cycles per image required for robust detection was 1.6, with an average intensity index of 122 (standard deviation 25) on a scale from 0 to 255. In addition to drusen identification, the platform reported the size of outlined lesions as well as the distance to the fovea and position angle. Corresponding location to the Wisconsin grading template can be calculated. Conclusions:A robust drusen analysis tool was developed and deployed to improve the accuracy and decrease the grading time for the purposes of clinical trials.

Keywords: age–related macular degeneration • drusen • 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.