June 2022
Volume 63, Issue 7
Open Access
ARVO Annual Meeting Abstract  |   June 2022
Characterising DARC (Detecting Apoptosing Retinal Cells) spots in glaucoma and healthy eyes
Author Affiliations & Notes
  • John Maddison
    Novai Ltd, United Kingdom
  • Soyoung Choi
    Glaucoma & Retinal Neurodegnrtn Res Grp, Imperial College, UCL IOO, Western Eye Hsp London, United Kingdom
    Novai Ltd, United Kingdom
  • M Francesca Cordeiro
    Glaucoma & Retinal Neurodegnrtn Res Grp, Imperial College, UCL IOO, Western Eye Hsp London, United Kingdom
    Novai Ltd, United Kingdom
  • Footnotes
    Commercial Relationships   John Maddison Novai Ltd, Code E (Employment), Novai Ltd, Code P (Patent); Soyoung Choi Novai Ltd, Code E (Employment); M Francesca Cordeiro Visufarma, Code C (Consultant/Contractor), Allergan, Code C (Consultant/Contractor), Aerie Pharmaceuticals, Code C (Consultant/Contractor), Novai Ltd, Code E (Employment), Santen, Code F (Financial Support), Thea, Code F (Financial Support), Heidelberg Engineering, Code F (Financial Support), Novartis, Code R (Recipient)
  • Footnotes
    Support  Wellcome Trust (grant no. WT099729)
Investigative Ophthalmology & Visual Science June 2022, Vol.63, 2033 – A0474. doi:
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      John Maddison, Soyoung Choi, M Francesca Cordeiro; Characterising DARC (Detecting Apoptosing Retinal Cells) spots in glaucoma and healthy eyes. Invest. Ophthalmol. Vis. Sci. 2022;63(7):2033 – A0474.

      Download citation file:


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

      ×
  • Supplements
Abstract

Purpose : Apoptosis is recognised as one of the earliest pathological processes in glaucoma. DARC technology is a retinal imaging technique the utilises an intravenous fluorescently labelled annexin 5 (ANX776) to identify cells in stress and apoptosis. We have previously shown, using a convolutional neural network (CNN)-aided algorithm, that the quantification of DARC spots (CNN DARC count) can predict progression of glaucoma up to 18 months ahead of OCT RNFL global changes. Here, we assess the qualitative characteristics of CNN-identified DARC spots by comparing their common descriptive measures (DM) in glaucoma and healthy eyes.

Methods : Anonymised retinal images were used from over-40year old glaucoma (n = 28 eyes) and healthy subjects (n = 68 eyes) from the Phase 2 DARC clinical trial (ISRCTN10751859). These subjects were recruited following informed consent, being obtained according to the Declaration of Helsinki and study approval by the Brent Research Ethics Committee. A previously described CNN-aided automated algorithm was used to identify DARC spots. Assessment of DM characteristics included analysis of mean, std (standard deviation), skew, kurtosis, cv (coefficient of variation), bvdist (Euclidean distance from the nearest blood vessel) and the 7 Hu moments from CNN-algorithm identified DARC spots averaged per eye. These DMs were analysed using single logistic regression (SLR) and bootstrapping (n = 1000) to classify glaucoma and healthy eyes.

Results : SLR identified cv as the best independent DM variable (AUC=0.75), with a log-likelihood ratio (LLR) p-value <0.0001. Then applying bootstrapping (n = 1000) to SLR to obtain confidence intervals gave an AUC of 0.75±0.07.

Conclusions : Our findings suggest that standard DMs of CNN-algorithm identified DARC spots (averaged per eye), may be used to classify glaucoma and healthy subjects. The cv DM appeared most promising using SLR, and can be easily applied to continuing DARC studies. In the future it may be possible to use an increased number of features to improve results. Future analysis of planned larger studies using DMs could further help improve disease classification, but these results support use of DM with CNN DARC for glaucoma diagnosis.

This abstract was presented at the 2022 ARVO Annual Meeting, held in Denver, CO, May 1-4, 2022, and virtually.

×
×

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.

×