Investigative Ophthalmology & Visual Science Cover Image for Volume 65, Issue 7
June 2024
Volume 65, Issue 7
Open Access
ARVO Annual Meeting Abstract  |   June 2024
Evaluation of Neovascular Regression in Rabbit Persistent Leak Model Using Deep Learning
Author Affiliations & Notes
  • Lisa Marie Hernandez-Denlinger
    Ophthalmology Discovery Research, AbbVie Inc, Irvine, California, United States
  • Footnotes
    Commercial Relationships   Lisa Hernandez-Denlinger None
  • Footnotes
    Support  None
Investigative Ophthalmology & Visual Science June 2024, Vol.65, 4950. doi:
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Lisa Marie Hernandez-Denlinger; Evaluation of Neovascular Regression in Rabbit Persistent Leak Model Using Deep Learning. Invest. Ophthalmol. Vis. Sci. 2024;65(7):4950.

      Download citation file:


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

      ×
  • Supplements
Abstract

Purpose : Accurate quantification of retinal neovascularization (NV) is critical to determining the effectiveness of therapeutic agents in pre-clinical models of neovascular age-related macular degeneration (nAMD). The Persistent Retinal Vascular Leak Model (PRVL), induced in rabbits through intravitreal injection (IVT) of DL-alpha-aminoadipic acid (DLAAA), serves as the gold standard animal model for studying drug efficacy for nAMD. The purpose of this study is to develop a deep learning (DL)-based algorithm to automatically segment and quantify the regression of new vessels following drug treatments.

Methods : High Resolution FITC (Fluorescein Isothiocyanate)-dextran fundus images were obtained from 38 PRVL-induced rabbits treated with AGN-888 (500 µg, 20 µg, and 3 µg) or Aflibercept (AFL) (20 µg and 3 µg), in comparison to control (PBS) group. The animals underwent biweekly imaging for the first 12 weeks, followed by every 4 weeks up to 32 weeks. A DL model with U-Net architecture is trained to segment and calculate vascular areas in the FITC-dextran images. The images were split into training set (n=100) and test set (n=50), with each image containing approximately 200 vessels. The Dice similarity coefficient (DSC) was used as a metric to evaluate the performance of DL segmentations.

Results : The DL model successfully segmented and identified the pathological neovessels from existing vasculature. The average DSC of neovessels were 0.99, 0.96 and 0.99 for AGN-888 500 µg, AFL 20 µg and control, respectively. Both AGN-888 500 µg and AFL 20 µg were observed to regress but AGN-888 showed higher efficacy (62% regression) compared to AFL 20 µg (24% regression) at 32 weeks. There was no statistically significant difference observed between AGN-888 and Aflibercept at lower concentrations.

Conclusions : DL-based quantification of retinal vasculature provides accurate and robust measurements of neovasculature and the existing vasculature area. This, in turn, enables the rapid and automated calculation of vessel regression, which is vital in assessing the efficacy of therapeutic interventions in neovascular age-related macular degeneration (nAMD) pre-clinical models.

This abstract was presented at the 2024 ARVO Annual Meeting, held in Seattle, WA, May 5-9, 2024.

×
×

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.

×