June 2020
Volume 61, Issue 7
Open Access
ARVO Annual Meeting Abstract  |   June 2020
Deep learning-based automated fluid quantification in clinical routine OCT images in neovascular AMD
Author Affiliations & Notes
  • Bianca S S. Gerendas
    Medical University of Vienna, Vienna, Austria
  • Amir Sadeghipour
    Medical University of Vienna, Vienna, Austria
  • Martin Michl
    Medical University of Vienna, Vienna, Austria
  • Thomas Alten
    Medical University of Vienna, Vienna, Austria
  • Wolf Buehl
    Medical University of Vienna, Vienna, Austria
  • Stefan Sacu
    Medical University of Vienna, Vienna, Austria
  • Hrvoje Bogunovic
    Medical University of Vienna, Vienna, Austria
  • Ursula Schmidt-Erfurth
    Medical University of Vienna, Vienna, Austria
  • Footnotes
    Commercial Relationships   Bianca S Gerendas, IDx (F), Novartis (C), Roche (C); Amir Sadeghipour, None; Martin Michl, None; Thomas Alten, None; Wolf Buehl, None; Stefan Sacu, Bayer (F), Bayer (C), Novartis (F), Novartis (C); Hrvoje Bogunovic, None; Ursula Schmidt-Erfurth, Boehringer Ingelheim (C), Carl Zeiss Meditec (C), Genentech (F), Novartis (C), Novartis (R), Novartis (F), Roche (C)
  • Footnotes
    Support  None
Investigative Ophthalmology & Visual Science June 2020, Vol.61, 2359. doi:
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Bianca S S. Gerendas, Amir Sadeghipour, Martin Michl, Thomas Alten, Wolf Buehl, Stefan Sacu, Hrvoje Bogunovic, Ursula Schmidt-Erfurth; Deep learning-based automated fluid quantification in clinical routine OCT images in neovascular AMD. Invest. Ophthalmol. Vis. Sci. 2020;61(7):2359.

      Download citation file:


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

      ×
  • Supplements
Abstract

Purpose : VIBES (Vienna Imaging Biomarker Eye Study) is a registry of real-world patient data (2007-2017) from the Department of Ophthalmology, Medical University of Vienna. To date, a total of 585,919 optical coherence tomography (OCT) scans of 38,295 patients have been collected. The purpose of this analysis was to quantify the amount of intra- and subretinal fluid before anti-vascular endothelial growth factor (VEGF) treatment initiation of patients with neovascular age-related macular degeneration (nAMD) and up to 5 years after their first anti-VEGF injection.

Methods : Data from two electronic patient record databases (7,832 and 6,677 patients), the hospital controlling information on treatments (42,965 procedures on 6,928 patients) and OCT images from a Zeiss Cirrus (24,473 patients) and a Heidelberg Spectralis database (13,822 patients) were analyzed. Matching all entries and filtering for AMD and at least one anti-VEGF injection for active nAMD led to 1138 eyes (898 Zeiss Cirrus, 240 Heidelberg Spectralis) with an available baseline OCT scan from 60 to 0 days before their first injection. 656 eyes had a follow-up OCT scan of 1 year, 408 of 2 years, 309 of 3, 221 of 4 and 175 of 5 years. Visual acuity (VA) at baseline and years 1-5, age, gender, OCT from baseline and follow-up and number of treatments in each year were included in the analysis. OCTs were automatically analyzed for central retinal thickness (CRT) and volume of intra- and subretinal fluid by a validated deep learning algorithm.

Results : Baseline age was 77.4 years (38% male/62% female). Automated segmentation failed in only 0.002% of visits. In mean, the baseline OCT scan was taken 11.2 days before the first injection. Mean baseline VA was 0.34 (Snellen 20/59). During follow-up it was 0.40 (20/50) at year 1, 0.39 (20/51) at year 2, 0.36 (20/55) at year 3, 0.33 (20/60) at year 4 and 0.32 (20/61) at year 5 receiving a mean of 4.26/3.08/2.78/2.55/2.45 injections per year. Mean CRT was 358µm at baseline and around 300µm at subsequent years. Mean total intraretinal fluid volume was 70nl at baseline and 35-51nl at subsequent years. Mean total subretinal fluid volume was 306nl at baseline and around 100nl at subsequent years.

Conclusions : Deep learning-based automated fluid quantification is well-suited to objectively, reliably and rapidly measure treatment response after anti-VEGF treatment and guide the clinical management in nAMD.

This is a 2020 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.

×