June 2021
Volume 62, Issue 8
Open Access
ARVO Annual Meeting Abstract  |   June 2021
Association between retinal thickness and retinal fluid volumes measured by deep learning in the HAWK & HARRIER trials
Author Affiliations & Notes
  • Maximilian Pawloff
    Medizinische Universitat Wien, Wien, Wien, Austria
  • Hrvoje Bogunovic
    Medizinische Universitat Wien, Wien, Wien, Austria
  • Zufar Mulyukov
    Novartis AG, Basel, Basel-Stadt, Switzerland
  • Daniel Lorand
    Novartis AG, Basel, Basel-Stadt, Switzerland
  • Ursula Schmidt-Erfurth
    Medizinische Universitat Wien, Wien, Wien, Austria
  • Footnotes
    Commercial Relationships   Maximilian Pawloff, None; Hrvoje Bogunovic, None; Zufar Mulyukov, Novartis AG (E); Daniel Lorand, Novartis AG (E); Ursula Schmidt-Erfurth, Boehringer Ingelheim (C), Genentech (C), Heidelberg Engineering (C), Kodiak (C), Novartis (C), Roche (C)
  • Footnotes
    Support  None
Investigative Ophthalmology & Visual Science June 2021, Vol.62, 2105. doi:
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Maximilian Pawloff, Hrvoje Bogunovic, Zufar Mulyukov, Daniel Lorand, Ursula Schmidt-Erfurth; Association between retinal thickness and retinal fluid volumes measured by deep learning in the HAWK & HARRIER trials. Invest. Ophthalmol. Vis. Sci. 2021;62(8):2105.

      Download citation file:


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

      ×
  • Supplements
Abstract

Purpose : To investigate whether central subfield retinal thickness (CSFT) values correlate with exudative activity of patients undergoing treatment with anti-VEGF substances (Brolucizumab and Aflibercept). Fluid measurements i.e. intraretinal fluid (IRF) and subretinal fluid (SRF) were obtained by deep learning based on spectral-domain optical coherence tomography (SD-OCT) scans.

Methods : We utilized a previously validated, fully automated deep learning approach based on convolutional neural networks to detect and quantify macular fluid in SD-OCT volumes (Cirrus, Spectralis, Topcon) from patients gathered from multicenter studies HAWK (1078) & HARRIER (739) with neovascular age-related macular disease. Volumes (nl) for IRF and SRF were measured at baseline and monthly under anti-VEGF therapy in the central mm. The dataset was analyzed using descriptive statistics. We summarized the association between fluid volumes and CSFT measured from ILM to outer retinal pigment epithelium (RPE) boundaries, using the Pearson’s r correlation coefficient.

Results : 41,906 SD-OCT volume scans underwent fluid volume analysis using deep learning. Detection level AUC values for IRF in the central millimeter were 0.849 in HAWK and 0.933 in HARRIER. For SRF, AUC values were 0.874 for HAWK and 0.871 for HARRIER. In HAWK, IRF volumes showed only a moderate association with CSFT at baseline (r=0.54) and an even weaker correlation between CSFT and IRF under therapy (r=0.44). Correlation of SRF and CSFT was very weak at baseline (r=0.29) and did not increase much under therapy (r=0.38). Consistently, in HARRIER, IRF volumes had a moderate correlation of r=0.62 with CSFT at baseline. Under therapy, association of CSFT and IRF was weak (r=0.34). Correlation between SRF and CSFT was very week at baseline (r=0.22) and increased only marginally under therapy (r=0.45).

Conclusions : In patients with nAMD, CSFT does not adequately represent exudative activity and IRF and SRF volumes are more clinically relevant particularly during therapy. Due to the limited correlation of CSFT with retinal fluid, our findings do not support the use of CSFT as the main anatomical measure for treatment guidance and disease management. In contrast quantitative deep learning methods allow for a precise determination of fluid volumes and may be used as an optimal clinical decision support tool in anti-VEGF therapy in patients with nAMD.

This is a 2021 ARVO Annual Meeting abstract.

×
×

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.

×