June 2017
Volume 58, Issue 8
Open Access
ARVO Annual Meeting Abstract  |   June 2017
Automatic detection of cones in multi-modal adaptive optics scanning light ophthalmoscope images of achromatopsia
Author Affiliations & Notes
  • David Cunefare
    Biomedical Engineering, Duke University, Durham, North Carolina, United States
  • Christopher S Langlo
    Ophthalmology, Medical College of Wisconsin, Milwaukee, Wisconsin, United States
  • Alfredo Dubra
    Ophthalmology, Stanford University, Stanford, California, United States
  • Joseph Carroll
    Ophthalmology, Medical College of Wisconsin, Milwaukee, Wisconsin, United States
  • Sina Farsiu
    Biomedical Engineering, Duke University, Durham, North Carolina, United States
    Ophthalmology, Duke University, Durham, North Carolina, United States
  • Footnotes
    Commercial Relationships   David Cunefare, None; Christopher Langlo, None; Alfredo Dubra, Athena Vision (C), Meira GTx (C), US Patent 8,226,236 (P); Joseph Carroll, AGTC (F), Meira GTX (C); Sina Farsiu, None
  • Footnotes
    Support  Foundation for Fighting Blindness, R01EY017607, P30EY001931, R01EY025231, U01EY025477, P30- EY005722
Investigative Ophthalmology & Visual Science June 2017, Vol.58, 300. doi:
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to Subscribers Only
      Sign In or Create an Account ×
    • Get Citation

      David Cunefare, Christopher S Langlo, Alfredo Dubra, Joseph Carroll, Sina Farsiu; Automatic detection of cones in multi-modal adaptive optics scanning light ophthalmoscope images of achromatopsia. Invest. Ophthalmol. Vis. Sci. 2017;58(8):300.

      Download citation file:


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

      ×
  • Supplements
Abstract

Purpose : To develop an automatic method for detecting cones in low-contrast “clinical grade” adaptive optics scanning light ophthalmoscope (AOSLO) images of subjects with achromatopsia (ACHM).

Methods : The automatic matched-filter based algorithm combined filter responses from simultaneously captured split detector and confocal AOSLO images to detect the cone locations. The parameters for our algorithm were trained on 40 pairs of images from 3 ACHM subjects, and the algorithm was validated on an additional 40 pairs from 3 previously unseen ACHM subjects. One-to-one matches between the automatic results and manual markings on the split detector images (made independently of the corresponding confocal images) were found in order to calculate measures of sensitivity and false discovery rate across the validation set. A second manual grader qualitatively compared the results of automated and manual cone detection in each image.

Results : Figure 1 shows a qualitative example of the automatic segmentation compared to expert manual marking. Of the 1871 cones that were manually detected (46.8 ± 23.8 cones per image) the automated method had an average sensitivity of 0.88 and false discovery rate of 0.20. The average computation time was 11 milliseconds per image.

Conclusions : There was an overall good agreement between automatic and manual grading. Manual grading of cone locations in AOSLO images of diseased eyes is difficult and subjective (Abozaid, et al. Adv Exp Med Biol, 854, 2016), and qualitative assessment by the second grader revealed that in several cases, cones missed by the grader were detected by the algorithm. These potential errors in manual grading negatively impacted the quantitative performance metric for the automatic method. Consensus of multiple graders will be used to improve accuracy of manual grading in future studies.

This is an abstract that was submitted for the 2017 ARVO Annual Meeting, held in Baltimore, MD, May 7-11, 2017.

 

Figure 1: a) Split detector AOSLO image of cones in an ACHM subject. Green (automatic) and yellow (manual) asterisks denote a correctly identified cone, cyan is a cone missed by the algorithm (false negative), and red is a location marked by the algorithm but not manually (false positive). Examples of false positives that are likely cones missed by the manual grader are pointed to with orange arrows. b) Corresponding confocal AOSLO image with the overlay of orange arrows from (a) pointing to potential cone structures.

Figure 1: a) Split detector AOSLO image of cones in an ACHM subject. Green (automatic) and yellow (manual) asterisks denote a correctly identified cone, cyan is a cone missed by the algorithm (false negative), and red is a location marked by the algorithm but not manually (false positive). Examples of false positives that are likely cones missed by the manual grader are pointed to with orange arrows. b) Corresponding confocal AOSLO image with the overlay of orange arrows from (a) pointing to potential cone structures.

×
×

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.

×