September 2016
Volume 57, Issue 12
ARVO Annual Meeting Abstract  |   September 2016
Assessment of the accuracy of outer iris boundary detections
Author Affiliations & Notes
  • Li Chen
    AMO Development, LLC, Milpitas, California, United States
  • Dimitri Chernyak
    AMO Development, LLC, Milpitas, California, United States
  • Footnotes
    Commercial Relationships   Li Chen, Abbott Medical Optics, Inc. (E); Dimitri Chernyak, Abbott Medical Optics, Inc. (E)
  • Footnotes
    Support  None
Investigative Ophthalmology & Visual Science September 2016, Vol.57, 5935. doi:
  • Views
  • Share
  • Tools
    • Alerts
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Li Chen, Dimitri Chernyak; Assessment of the accuracy of outer iris boundary detections. Invest. Ophthalmol. Vis. Sci. 2016;57(12):5935.

      Download citation file:

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

  • Supplements

Purpose : The accuracy of outer iris boundary (OIB) detections in eye images is critical for iris registration. Compared to the sharp edge of pupil boundary, OIB detection is more challenging for image processing. We investigate whether the OIB detections made by image processing algorithm on eye images is equivalent to the OIB detections by human eyes.

Methods : Scotopic eye images of 62 eyes were captured by iDesign Wavefront Studio system (AMO Development, LLC). The images were analyzed to locate OIB with the image processing algorithm implemented on iDesign® system. 5 trained people participated as observers. Each observer used a software tool to visually inspect the accuracy of OIB placement in the processed iDesign® images. If the OIB location error was found by the observer, he/she could use this tool to place a new circle for the OIB into the raw image by changing the OIB center position (X, Y) and radius (R). Repeatabilty of the OIB adjustment was conducted by each observer on one image. A matched pair metric of equivalence was applied to the matched pair analysis for OIB detection difference made by the image processing algorithm and the mean made by the 5 observers. The metric specified that if the 95% confidence limits lie within the measurement’s tolerance range (-50μm, 50μm), then the two measurement distributions from the production algorithm and the observers were equivalent to that confidence level.

Results : On average, observers made the OIB adjustments with the 62 images mainly on the horizontal center position (Delta X) and the radius of the OIB circle (Delta R). The repeatibility test shows the OIB adjustment on iDesign® system images made by human eye is repeatable. Table1 shows the statistical results with the 62 iDesign® images. The standard deviations from the OIB adjustments show that the OIB placements made by the 5 observers were within the measurement’s tolerance range. The mean difference and the 95% confidence interval limits also lie within the tolerance range . Therefore, the OIB placements made by the image processing algorithm and human observers with the 62 iDesign® system images are equivalent.

Conclusions : The quantitative assessment of the accuracy of OIB detections made by the image processing on iDesign® system was evaluated by 5 observers. With the 62 iDesign® system images, the OIB detection made by the image processing algorithm on iDesign® system is equivalent to the OIB detection by human eyes.

This is an abstract that was submitted for the 2016 ARVO Annual Meeting, held in Seattle, Wash., May 1-5, 2016.



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.