Purchase this article with an account.
Veronika Röggla, Gregor Sebastian Reiter, Philipp Fuchs, Oliver Leingang, Leonard Mana Coulibaly, Hrvoje Bogunovic, Daniel Barthelmes, Ursula Schmidt-Erfurth; Spatial and temporal distribution of recurrent fluid in neovascular AMD quantified with a deep learning algorithm. Invest. Ophthalmol. Vis. Sci. 2022;63(7):3019 – F0289.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
The aim of the study is to investigate the quantity and distribution of fluid in neovascular age-related macular degeneration (nAMD) and the difference between the initial, treatment-naïve, presentation, at the beginning and the end of a treatment break.
Using data from one site of the Fight Retinal Blindness (FRB) Registry (University of Zurich, Switzerland), the distribution of retinal fluid in treatment-naïve nAMD patients compared to retinal fluid recurrence has been studied in a longitudinal manner. Study eye eligibility required treatment-naïve nAMD, receiving at least three anti-VEGF injections within the first six months of treatment, followed by a treatment pause of at least six months and fluid recurrence afterwards. To quantify fluid, a verified deep learning algorithm was used (Vienna Fluid Monitor, RetInSight, Vienna, Austria). Fluid volumes and proportions were compared between different time points using Friedman test and Wilcoxon signed rank test, respectively.
Fifty-five eyes of 55 patients with the mean age of 76.86 ± 6.04years at baseline matched the inclusion criteria. A median of 12.66 (IQR 11) injections and 18.96 (IQR 22) months of treatment were needed to enable a treatment pause of at least six months. Visual acuity was 0.57 logMar at baseline, 0.47 logMar at start of the break and 0.48 logMar at the end of the break. At baseline, SRF showed a median of 55.29nl (IQR 243.79nl), IRF a median of 1.55nl (IQR 133.37nl) and PED a median of 0nl (IQR 520.73nl) in the central 6 millimeters. At disease recurrence, SRF showed a median of 8.07nl (IQR 86.29nl), IRF a median of 4.88nl (IQR 24.85nl) and PED a median of 0nl (IQR 551.61nl) in the central 6 millimeters (Figure 1). From baseline to start of the break, SRF and IRF decreased significantly (p<0.01) and PED did not change significantly (p=0.215). At baseline, the proportions of IRF, SRF and PED were 21%, 53% and 26%, respectively, with no significant shift except SRF at end of break to 27%,41% and 32% (p=0.123, p=0.029 and p=0.374), respectively.
Retinal fluid proportions were distributed predominantly in the subretinal compartment in treatment-naïve patients, whereas it was spread more equally between all fluid compartments after a treatment break. These morphological differences may contribute to our understanding of disease activity.
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