June 2015
Volume 56, Issue 7
ARVO Annual Meeting Abstract  |   June 2015
Photograph Reading Center Front-end Automated Fundus Image Analysis for Clinical Trials
Author Affiliations & Notes
  • Brian Madow
    Ophthalmology, University of South Florida, Tampa, FL
  • Footnotes
    Commercial Relationships Brian Madow, None
  • Footnotes
    Support None
Investigative Ophthalmology & Visual Science June 2015, Vol.56, 5255. doi:
  • Views
  • Share
  • Tools
    • Alerts
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Brian Madow; Photograph Reading Center Front-end Automated Fundus Image Analysis for Clinical Trials. Invest. Ophthalmol. Vis. Sci. 2015;56(7 ):5255.

      Download citation file:

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

  • Supplements

To design and implement automated query analysis of the submitted fundus photograph images to the Photograph Reading Center (PRC) for clinical trials by evaluating the characteristics of the digital data and perform automatic check against the required parameters. Currently the submitted images to the PRC are evaluated manually, which is time and labor consuming process. We aimed to increase the accuracy, decrease the time for the processing and increase the effectiveness at the busy PRC.


Fast programing languages were used to develop algorithms and friendly graphical user interface (GUI). The software was tested on computer with Pentium quad processor running Windows 7 operating system. Small set of 25 digital images was initially used to test the principle and adjust the algorithms. Set of 100 consecutively received fundus photographic images at the PRC were evaluated for accepted file format, file location, compression, file size, resolution, image name, identifiable subject information, image quality, and image color profile. Additional algorithm was developed to extract and display the coded into the file information about the name of the study, the clinical site, the subject code, the visit number, the image type, the eye (right or left) and the date when the image was taken. Supplementary module was added to allow for operator input for other image characteristics pertinent to the clinical trial that cannot be automated.


The software performance was checked for accuracy against manually evaluating the images by an operator. The test results yielded 100 % accuracy from automated output. Images were retested repeatedly 3 times and the accuracy was found to be reproducible 100%. The time for performing the automated testing was found to be less than 1 sec., which is impossible to achieve by manual evaluation of the images. The extracted information from the file was found to be 100% accurate, when encoded as required.


Robust, reliable, automated software platform was created, evaluated, validated and implemented at the Uveitis Reading Center for assessment of vitreous haze grade from fundus photographs. The software is applicable to fundus images of any type and can be employed for use in other Reading centers or for other studies of fundus images.


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.