Purpose: To predict improvement of best-corrected visual acuity (BCVA) 1 year after pars plana vitrectomy for epiretinal membrane (ERM) using artificial intelligence methods on optical coherence tomography B-scan images. Methods: Four hundred and eleven (411) patients with Stage II ERM were divided in a group improvement (IM) (≥15 ETDRS letters of VA recovery) and a group no improvement (N-IM) (<15 letters) according to 1-year VA improvement after 25-G pars plana vitrectomy with internal limiting membrane peeling. Primary outcome was the creation of a deep learning classifier (DLC) based on optical coherence tomography B-scan images for prediction. Secondary outcome was assessment of the influence of various clinical and imaging predictors on BCVA improvement. Inception-ResNet-V2 was trained using standard augmentation techniques. Testing was performed on an external data set. For secondary outcome, B-scan acquisitions were analyzed by graders both before and after fibrillary change processing enhancement. Results: The overall performance of the DLC showed a sensitivity of 87.3% and a specificity of 86.2%. Regression analysis showed a difference in preoperative images prevalence of ectopic inner foveal layer, foveal detachment, ellipsoid zone interruption, cotton wool sign, unprocessed fibrillary changes (odds ratio = 2.75 [confidence interval: 2.49–2.96]), and processed fibrillary changes (odds ratio = 5.42 [confidence interval: 4.81–6.08]), whereas preoperative BCVA and central macular thickness did not differ between groups. Conclusion: The DLC showed high performances in predicting 1-year visual outcome in ERM surgery patients. Fibrillary changes should also be considered as relevant predictors.

To use artificial intelligence to identify imaging biomarkers for anatomic and functional progression of lamellar macular hole (LMH) and elaborate a deep learning (DL) model based on OCT and OCT angiography (OCTA) for prediction of visual acuity (VA) loss in untreated LMHs. Multicentric retrospective observational study. Patients aged >18 years diagnosed with idiopathic LMHs with availability of good quality OCT and OCTA acquisitions at baseline and a follow-up >2 years were recruited. A DL model based on soft voting of 2 separate models (OCT and OCTA-based respectively) was trained for identification of cases with VA loss >5 ETDRS letters (attributable to LMH progression only) during a 2-year follow-up. Biomarkers of anatomic and functional progression of LMH were evaluated with regression analysis, feature learning (support vector machine [SVM] model), and visualization maps. Ellipsoid zone (EZ) damage, volumetric tissue loss (TL), vitreopapillary adhesion (VPA), epiretinal proliferation, central macular thickness (CMT), parafoveal vessel density (VD) and vessel length density (VLD) of retinal capillary plexuses, choriocapillaris (CC), and flow deficit density (FDD). Functionally progressing LMHs (VA-PROG group, 41/139 eyes [29.5%]) showed higher prevalence of EZ damage, higher volumetric TL, higher prevalence of VPA, lower superficial capillary plexus (SCP), VD and VLD, and higher CC FDD compared with functionally stable LMHs (VA-STABLE group, 98/139 eyes [70.5%]). The DL and SVM models showed 92.5% and 90.5% accuracy, respectively. The best-performing features in the SVM were EZ damage, TL, CC FDD, and parafoveal SCP VD. Epiretinal proliferation and lower CMT were risk factors for anatomic progression only. Deep learning can accurately predict functional progression of untreated LMHs over 2 years. The use of AI might improve our understanding of the natural course of retinal diseases. The integrity of CC and SCP might play an important role in the progression of LMHs. The authors have no proprietary or commercial interest in any materials discussed in this article.