Abstract
Purpose :
To analyze short-term changes of mean photoreceptor thickness (PRT) on the ETDRS-Grid after vitrectomy and membrane peeling in patients with epiretinal membrane (ERM) using deep-learning.
Methods :
Forty-eight patients with idiopathic ERM were included in this prospective study. Healthy fellow eyes served as control. Study examinations comprised best-corrected visual acuity (BCVA) and spectral-domain optical coherence tomography before surgery, 1 week, 1 month and 3 months after surgery. Mean PRT was assessed using an automated algorithm and correlated with BCVA and central retinal thickness (CRT).
Results :
Regarding PRT, a significant decrease from baseline to week 1 in the central 1mm (38.23µm±6.88µm to 32.13µm±5.61µm, p<0.001), 3mm (34.52µm±4.09µm to 29.78µm±4.24µm, p<0.001) and 6mm (32.40µm±3.69µm to 28.17µm±3.72µm, p<0.001) disc area as well as a significant increase at month 3 in the 6mm (32.40µm±3.69µm to 32.99µm±2.60µm, p=0.031) disc area of the study eye compared to the fellow eye (at baseline 40.82µm±5.62µm, 36.29µm±4.61µm and 33.46µm±4.68µm in the 1mm, 3mm and 6mm disc area, respectively; at week 1 40.35µm±4.91µm, 36.79µm±3.63µm and 33.97µm±4.12µm in the 1mm, 3mm and 6mm disc area, respectively; at month 3 33.07µm±4.15µm in the 6mm disc area) were observed. Changes in PRT in the central 1mm, 3mm and 6mm disc area after 1 week correlated negatively with the change in BCVA (0.30±0.24 logMar to 0.25±0.17 logMar; rs=-0.350, p=0.034; rs=-0.337, p=0.041 and rs=-0.343, p=0.038, respectively) and CRT (449.90µm±78.97µm to 462.20µm±52.53µm; rs=-0.604, p<0.001; rs=-0.399, p=0.013 and rs=-0.344, p=0.034, respectively). Visual acuity increased significantly from baseline (0.30±0.24 logMar) to month 3 (0.15±0.16 logMar, p<0.001).
Conclusions :
Application of an automated algorithm across a 6mm grid with different sections allows for a comprehensive evaluation of the photoreceptor layer in the macula. Early changes of PRT allow prediction about postoperative functional and even morphological outcomes.
This abstract was presented at the 2022 ARVO Annual Meeting, held in Denver, CO, May 1-4, 2022, and virtually.