Accuracy of retinal layers optical coherence tomography automated segmentation before and after epiretinal peeling

Purpose To access the accuracy of Spectral Domain – Optical Coherence Tomography (SD-OCT) retinal layers automated segmentation before and after idiopathic epiretinal membrane (ERM) peeling. Methods Retrospective observational study that included 37 eyes of 37 patients with ERM. OCT scans before and after peeling were obtained with Spectralis Heidelberg Engineering©. Two independent observers determined the accuracy of the automated retinal layer segmentation in the 3 mm central area. Manual correction of the automated segmentation was made as needed by the two observers and the new measured individual layers central thickness (CT) was compared to the one given by automated segmentation. Results Agreement between observers was perfect for the exams that needed manual segmentation correction. The agreement was moderate (κ = 0.51) for outer plexiform layer (OPL) before surgery and good to perfect (κ > 0.62) for the other layers before and after surgery. Preoperatively, 81.8% of the exams needed manual correction. The obtained CT was different (p < 0.05) before and after manual correction for central macular thickness, retinal nerve fibre layer (RNFL), ganglion cell layer, inner nuclear layer (INL), OPL and outer nuclear layer; no difference was noted in the inner plexiform layer and outer retinal layers (between external limiting and Bruch membranes). Postoperatively, 29.7% of the exams needed manual correction. There was a significant difference in the CT obtained with automated segmentation versus manual correction only in RNFL and INL. Conclusions OCT automated segmentation is not accurate in the internal retinal layers of patients with ERM before surgery, probably due to the much altered structure of these layers. After the surgery its accuracy is better. We recommend the verification of the automated segmentation in patients with ERM before surgery.