UCL Discovery Stage
UCL home » Library Services » Electronic resources » UCL Discovery Stage

A CNN-aided method to predict glaucoma progression using DARC (Detection of Apoptosing Retinal Cells)

Normando, EM; Yap, TE; Maddison, J; Miodragovic, S; Bonetti, P; Almonte, M; Mohammad, N; ... Cordeiro, MF; + view all (2020) A CNN-aided method to predict glaucoma progression using DARC (Detection of Apoptosing Retinal Cells). Expert Review of Molecular Diagnostics , 20 (7) pp. 737-748. 10.1080/14737159.2020.1758067. Green open access

[thumbnail of Accepted manuscript (final).pdf]
Preview
Text
Accepted manuscript (final).pdf - Accepted Version

Download (2MB) | Preview

Abstract

Artificial intelligence in ophthalmology has become an important analytical aid in retinal imaging, being frequently advocated in managing blinding conditions. One such condition is glaucoma - the leading cause of global irreversible blindness, affecting over 60.5 million people worldwide. A key objective in glaucoma research over the last few years is to find a biomarker to identify those at risk of rapid progression and blindness. Recently, a novel method aimed at visualising apoptotic cells in the retina in humans called DARC (Detection of Apoptosing Retinal Cells) has been reported. The molecular marker used in the technology is fluorescently labelled annexin A5, which has a high affinity for phosphatidylserine exposed on the surface of cells undergoing stress and in early stages of apoptosis. The published Phase 1 results suggested that the number of DARC positively stained cells seen in a retinal fluorescent image could be used to assess glaucoma disease activity, but also correlated with future glaucoma disease progression, albeit in small patient numbers (n=16). DARC has now been tested in even more subjects (n=120) in a Phase 2 clinical trial (ISRCTN10751859). We describe here an automatic method of DARC spot detection which was developed using a CNN which was trained and tested on the control cohort of subjects in the Phase 2 DARC trial. The CNN algorithm was found to have a 97.0% accuracy, 91.1% sensitivity and 97.1% specificity to spot detection when compared to manual grading of 50% controls. Subsequently, the CNN algorithm was tested on glaucoma patients in the trial, using gold standard optical coherence tomography (OCT) global rates of progression (retinal nerve fibre layer at 3.5 ring) eighteen months after their assessment with DARC. Those patients with a significant (p<0.05) negative slope were defined as progressing compared to those without who were defined as stable. The CNN algorithm had a sensitivity of 85.7% and specificity of 91.7% to glaucoma progression, with an AUC of 0.89. Finally, the CNN algorithm was found to show a significantly (p=0.0044) greater number of DARC positively stained cells in the progressing compared to stable glaucoma groups. This paper describes the successful use of a CNN-aided algorithm which automates detection of apoptosis with DARC enabling prediction of glaucoma progression 18 months later. We believe this method provides an automated and objective biomarker with potentially widespread clinical applications.

Type: Article
Title: A CNN-aided method to predict glaucoma progression using DARC (Detection of Apoptosing Retinal Cells)
Location: England
Open access status: An open access version is available from UCL Discovery
DOI: 10.1080/14737159.2020.1758067
Publisher version: https://doi.org/10.1080/14737159.2020.1758067
Language: English
Additional information: This version is the author accepted manuscript. For information on re-use, please refer to the publisher’s terms and conditions.
UCL classification: UCL
UCL > Provost and Vice Provost Offices > School of Life and Medical Sciences
UCL > Provost and Vice Provost Offices > School of Life and Medical Sciences > Faculty of Brain Sciences
UCL > Provost and Vice Provost Offices > School of Life and Medical Sciences > Faculty of Brain Sciences > Institute of Ophthalmology
URI: https://discovery-pp.ucl.ac.uk/id/eprint/10095650
Downloads since deposit
4,608Downloads
Download activity - last month
Download activity - last 12 months
Downloads by country - last 12 months

Archive Staff Only

View Item View Item