July 2019
Volume 60, Issue 9
Open Access
ARVO Annual Meeting Abstract  |   July 2019
Fully-Automated Estimation of Cone Metrics in Adaptive Optics Retinal Images
Author Affiliations & Notes
  • Robert F Cooper
    Psychology, University of Pennsylvania, Philadelphia, Pennsylvania, United States
    Scheie Eye Institute, Ophthalmology, University of Pennsylvania, Philadelphia, Pennsylvania, United States
  • Geoffrey K Aguirre
    Neurology, University of Pennsylvania, Philadelphia, Pennsylvania, United States
  • Jessica Ijams Wolfing Morgan
    Scheie Eye Institute, Ophthalmology, University of Pennsylvania, Philadelphia, Pennsylvania, United States
    Center for Advanced Retinal and Ocular Therapeutics, Ophthalmology, University of Pennsylvania, Philadelphia, Pennsylvania, United States
  • Footnotes
    Commercial Relationships   Robert Cooper, FourPi Innovations (P), MCW (C); Geoffrey Aguirre, None; Jessica Morgan, AGTC (F), US Patent 8226236 (P)
  • Footnotes
    Support  NIH R01EY028601, NIH U01EY025477, Research to Prevent Blindness, Foundation Fighting Blindness, the F. M. Kirby Foundation, and the Paul and Evanina Mackall Foundation Trust
Investigative Ophthalmology & Visual Science July 2019, Vol.60, 1427. doi:
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Robert F Cooper, Geoffrey K Aguirre, Jessica Ijams Wolfing Morgan; Fully-Automated Estimation of Cone Metrics in Adaptive Optics Retinal Images. Invest. Ophthalmol. Vis. Sci. 2019;60(9):1427.

      Download citation file:


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

      ×
  • Supplements
Abstract

Purpose : Automated image processing methods for adaptive optics scanning laser ophthalmoscopes (AOSLO) have enabled the fully objective creation of retinal images. However, quantitative analyses of the cone mosaic typically rely on subjectively selected regions of interest (ROI) and semi-automatically identified cone locations. Here, we assess a novel fully automated algorithm for determining pointwise inter-cell distance (ICD) and cone density.

Methods : We obtained images of the photoreceptor mosaic from 14 eyes of 9 subjects without retinal pathology at 2 time points using an AOSLO. Images from each time point and subject were randomly assigned to either test or validation datasets. To determine an image’s ICD and density, first the discrete Fourier transform (DFT) of the image was calculated. The radial average of the DFT was analyzed using a multi-scale fit-based algorithm that first established a rough estimate of the modal spacing, and used that estimate as an initial condition for fitting a piecewise function that ultimately was used to find the precise modal spacing. We then converted the modal spacing to ICD and cone density by assuming a triangularly packed mosaic. Modal ring distinctiveness, defined as its peak amplitude normalized by the residual amplitude, was used to establish measurement quality. The algorithm’s consistency was assessed between the two datasets, and accuracy was evaluated by comparing the results from the validation dataset against those determined from directly identified cones. Finally, the algorithm was extended to extract pointwise ICD and density from montages by calculating modal spacing over an overlapping grid of ROIs in a montage.

Results : The percent differences of DFT-derived ICD and density between the test and validation datasets were 5±5% and 9±10%, respectively (mean±SD), consistent with the differences in directly determined ICD and density (2±3%, 4±6%). DFT-derived ICD from the validation dataset was 0.04±0.37 µm greater than the corresponding directly determined ICD. Estimated cone density was 1,000±4,200 cones/mm2 less than the directly determined cone density. We successfully obtained pointwise ICD and cone density measurements across full montages.

Conclusions : We successfully created an accurate, repeatable, and fully automated algorithm for determining ICD and density in both individual ROIs and across entire montages.

This abstract was presented at the 2019 ARVO Annual Meeting, held in Vancouver, Canada, April 28 - May 2, 2019.

×
×

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.

×