Purpose: To develop a deep learning model classifying the laterality of optical coherence tomography (OCT) images. Methods: The study included two-dimensional OCT images (horizontal/vertical macular section) from Seoul National University Hospital. A deep learning model based on ResNet-18 was developed and trained to classify whether OCT images were horizontal or vertical sections and to predict the laterality of the images. Analysis of the results included calculating a mean area under the receiver operating characteristic curve (AUROC) and evaluating accuracy, specificity, and sensitivity. Gradient-weighted class activation for mapping visualization highlighted critical regions for classification. Results: A total of 5,000 eyes of 2,500 patients (10,000 images) was included in the development process. The test dataset consisted of 1,000 eyes of 500 patients (590 eyes without macular abnormalities, 208 epiretinal membranes, 111 age-related macular degenerations, 56 central macular edemas, 23 macular holes, and 12 other macular abnormalities). The deep learning model predicted the OCT section of the eyes in the test dataset with a mean AUROC of 0.9967. The accuracy, sensitivity, and specificity were 0.9835, 0.9870, and 0.9800, respectively. The model predicted the laterality of the eyes in horizontal OCT images with a mean AUROC of 1.0000. The accuracy, sensitivity, and specificity were 0.9970, 1.0000, and 0.9940, respectively. Using vertical OCT images, deep learning models failed to demonstrate any predictive performance in laterality classification. Conclusions We developed a deep learning model to classify the horizontal/vertical sections of OCT images and predict the laterality of horizontal OCT images with high accuracy, sensitivity, and specificity.

Background: To investigate the association of glycemic control and retinal microvascular changes in patients with type 2 diabetes mellitus (T2DM) without diabetic retinopathy (DR). Methods: This retrospective, observational, cohort study included patients with T2DM without DR. The patients were categorized into intensive control (IC; mean glycosylated hemoglobin [HbA1c] ≤7.0%) and moderate control (MC; mean HbA1c >7.0%) groups. Optical coherence tomography (OCT) and swept-source OCT angiography (OCTA) image parameters were compared between three groups, including healthy controls. Results: In total, 259 eyes of 259 participants (88 IC, 81 MC, and 90 controls) were included. The foveal avascular zone area was significantly larger in the MC group than IC and control groups (all P<0.05). The IC group had lower vessel density in the superficial retinal layer and deep retinal layer than the controls (all P<0.05). The choriocapillaris (CC) flow deficit (FD) was significantly greater in the MC group than in the IC and control groups (18.2%, 16.7%, and 14.2%, respectively; all P<0.01). In multivariate regression analysis, CC-FD was associated with the mean HbA1c level (P=0.008). There were no significant differences in OCT parameters among the groups. Conclusion OCTA revealed that early CC impairment is associated with HbA1c levels; the CC changes precede clinically apparent DR. The OCTA parameters differed among the groups according to the degree of glycemic control. Our results suggest that microvascular changes precede DR and are closely related to glycemic control.

Purpose: To report a case of full-thickness macular hole (FTMH) formation secondary to demarcation laser retinopexy in a retinal break with localized retinal detachment patient.Case summary: A 59-year-old male visited our clinic with ocular discomfort in both eyes. Uncorrected visual acuity (UCVA) was 0.63 in right eye. Large retinal break in 1:30 o/c, localized retinal detachment and laser marking scars all around the right eye were found in fundoscopic exam. Posterior-vitreous detachment or vitreomacular traction was not observed in optical coherence tomography (OCT). Demarcation laser retinopexy was performed around the margin of retinal detachment and peripheral degenerative lesions. Three months after demarcation laser retinopexy, UCVA in right eye of the patient was decreased to 0.16 and full thickness macular hole was observed on OCT examination. Pars planar vitrectomy, internal limiting membrane peeling, and SF6 gas tamponade were performed in the right eye. One month after the surgery, closure of FTMH was observed. Three months after surgery, there were no recurrence of FTMH in the right eye. Conclusions Demarcation laser photocoagulation of localized retinal detachments may predispose to FTMH formation. Even though it can be occurred rarely, follow-up check-up is necessary in consideration of the possibility of FTMH, which can cause serious visual loss.

Purpose: To develop a deep learning model that can predict the axial lengths of eyes using ultra-widefield (UWF) fundus photography. Methods: We retrospectively enrolled patients who visited the ophthalmology clinic at the Seoul National University Hospital between September 2018 and December 2021. Patients with axial length measurements and UWF images taken within 3 months of axial length measurement were included in the study. The dataset was divided into a development set and a test set at an 8:2 ratio while maintaining an equal distribution of axial lengths (stratified splitting with binning). We used transfer learning-based on EfficientNet B3 to develop the model. We evaluated the model’s performance using mean absolute error (MAE), R-squared (R2), and 95% confidence intervals (CIs). We used vanilla gradient saliency maps to illustrate the regions predominantly used by convolutional neural network. Results: In total, 8,657 UWF retinal fundus images from 3,829 patients (mean age, 63.98 ±15.25 years) were included in the study. The deep learning model predicted the axial lengths of the test dataset with MAE and R2 values of 0.744 mm (95% CI, 0.709–0.779 mm) and 0.815 (95% CI, 0.785–0.840), respectively. The model’s accuracy was 73.7%, 95.9%, and 99.2% in prediction, with error margins of ±1.0, ±2.0, and ±3.0 mm, respectively. Conclusions We developed a deep learning-based model for predicting the axial length from UWF images with good performance.

PURPOSE: To evaluate the 1-year results of vitrectomy performed in combination with intraoperative dexamethasone implant for tractional and nontractional refractory diabetic macular edema (DME). METHODS: Thirteen eyes from 13 subjects who were diagnosed with tractional DME and 17 eyes from 17 subjects who were diagnosed with nontractional refractory DME underwent vitrectomy and dexamethasone implant injection. Changes in best-corrected visual acuity (BCVA) and central macular thickness (CMT) during the one year following vitrectomy were evaluated in each group. Additionally, changes in intraocular pressure and other complications were investigated postoperatively. RESULTS: In eyes with tractional DME, a statistically significant improvement in BCVA was noted at 3, 6, and 12 months, and a statistically significant improvement in CMT was noted at 1, 3, 6, and 12 months from baseline after vitrectomy (p < 0.05). In eyes with nontractional refractory DME, a statistically significant improvement in BCVA was noted at 12 months, but there were no significant improvements in CMT despite the tendency to decrease from baseline. Sixteen (53.3%) of the 30 eyes included in this study showed intraocular pressure elevation, which was addressed using antiglaucoma medication, and there were no other severe complications. CONCLUSIONS: Vitrectomy combined with intraoperative dexamethasone implant may be safe and effective in treating DME, especially tractional DME. In this study, patients with nontractional DME required more additional treatments and time for anatomical and functional improvement compared to patients with tractional DME.
Dexamethasone ; Humans ; Intraocular Pressure ; Macular Edema ; Traction ; Visual Acuity ; Vitrectomy