Search

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 investigate the frequency of optic disc hyperfluorescence (ODH) and their association with macular and optic nerve head thickness in patients with active uveitis.Methods: This retrospective cross-sectional study included 37 eyes of 30 patients with active noninfectious uveitis who underwent spectral domain-optical coherence tomography (OCT) and fluorescein angiography. OCT parameters were compared between baseline and post-treatment according to the ODH.Results: ODH was observed clearly in 22 eyes (64.7%) and equivocally in five eyes (14.7%) at baseline. In subgroup analysis without clinically significant optic disc edema, central subfield thickness, peripapillary retinal nerve fiber layer thickness, disc rim area, and disc area were 262.2 μm, 110.9 μm, 1.49 mm2, and 2.02 mm2 at baseline and significantly decreased after treatment (249.5 μm, 102.1 μm, 1.34 mm2, and 1.95 mm2, respectively) in eyes with ODH (p = 0.044, p < 0.001, p = 0.001, p = 0.024, respectively). However, there was no significant changes in the eyes without ODH (all p > 0.05).Conclusions: There was a difference in the thickness reduction of the macula and optic nerve head according to the ODH. Baseline ODH should be regarded as a sign of ocular inflammation and carefully monitored.

We aimed to develop a deep learning model for detecting and localizing retinal breaks in ultrawidefield fundus (UWF) images.We retrospectively enrolled treatment-naive patients diagnosed with retinal break or rhegmatogenous retinal detachment and who had UWF images. The YOLO v3 architecture backbone was used to develop the model, using transfer learning. The performance of the model was evaluated using per-image classification and per-object detection.Overall, 4,505 UWF images from 940 patients were used in the current study. Among them, 306 UWF images from 84 patients were included in the test set. In per-object detection, the average precision for the object detection model considering every retinal break was 0.840. With the best threshold, the overall precision, recall, and F1 score were 0.6800, 0.9189, and 0.7816, respectively. In the per-image classification, the model showed an area under the receiver operating characteristic curve of 0.957 within the test set. The overall accuracy, sensitivity, and specificity in the test data set were 0.9085, 0.8966, and 0.9158, respectively.The UWF image-based deep learning model evaluated in the current study performed well in diagnosing and locating retinal breaks.

Purpose: To investigate the role of ultra-widefield fluorescein angiography (UWFA) for monitoring the therapeutic response to adalimumab in patients with Behcet’s uveitis.Methods: Patients with Behcet’s uveitis treated with adalimumab for ≥ 30 weeks were included. The best-corrected visual acuity (BCVA), grading of anterior chamber cell, grading of vitreous haze, and UWFA findings were assessed at baseline, and at 6, 14, and 30 weeks from the start of adalimumab treatment.Results: A total of 32 eyes of 17 patients were included. There were significant decreases in BCVA and UWFA scores at 6, 14, and 30 weeks after adalimumab administration. UWFA scores for vascular and capillary leakage were decreased at 30 weeks (p = 0.032, p = 0.004), while the UWFA scores for disc hyperfluorescence persisted (p = 0.083). Although vascular and capillary leakage at the posterior pole was significantly decreased at 6, 14, and 30 weeks, there was no significant change of the score for peripheral vascular and capillary leakage from baseline to 6 weeks, from 6 to 14 weeks, and from 14 to 30 weeks.Conclusions: UWFA can be used to evaluate the treatment response of adalimumab in Behcet’s uveitis. The vascular leakage pattern varies depending on the location of retinal vessels.

To investigate longitudinal changes in the retinal and choroidal microstructure of the macula in patients with retinitis pigmentosa (RP).Retrospective, observational cohort study.A total of 69 patients with RP and 69 age- and sex-matched controls who underwent optical coherence tomography (OCT) over a 4-year follow-up period were included. The severity of RP was classified into 3 stages according to the integrity of the inner segment ellipsoid zone. The retinal and choroidal layers were segmented manually from OCT images. The areas of retinal pigment epithelium (RPE) atrophy and choroidal vascular index (CVI) were also analyzed. Longitudinal changes in the OCT parameters were compared among the groups.Significant decreases (median [interquartile range]) in the thickness of the ganglion cell inner plexiform layer (GCIPL; -1.04 [-2.41 to -0.17]), outer nuclear layer (ONL; -1.44 [-1.86 to -0.28]), and inner segment ellipsoid (ISE; -0.74 [-1.33 to -0.49]) at the moderate stage and retinal nerve fiber layer (RNFL; -1.49 [-2.08 to -0.66]) and GCIPL (0.58 [-1.79 to 0.06]) at the advanced stage were observed. Choroidal thickness decreased significantly from -7.62 to -9.40 μm per year at all stages. RPE atrophy and CVI reduction were observed at the advanced stage. There was no change in the control group.ONL and GCIPL thicknesses decreased at the moderate and advanced stages of RP; RNFL thickness decreased only at the advanced stage; and choroidal thickness decreased continuously. In addition, RPE atrophy and CVI reduction were prominent at the advanced stage. These results indicate that there is a temporal variation in the damage of each retinal layer and the choroid in RP patients.

Purpose To investigate the association between the presence of a cilioretinal artery (CRA) and advanced age-related macular degeneration (AMD), including the prevalence of choroidal neovascularization (CNV) and geographic atrophy (GA). Design Retrospective cross-sectional study. Methods Setting: Single center Patient population: A total of 738 patients with AMD who underwent optical coherence tomography angiography (OCTA) were included in the study. Fundus photographs were reviewed to determine the presence of the CRA. Intervention/observation procedures: In patients with a unilateral CRA, paired tests were performed between eyes with and without the CRA to compare AMD severity and prevalence of CNV and GA. Main outcomes: AMD stage, and prevalence of CNV and GA. Macular vasculature, including vessel density, perfusion density, and foveal avascular zone, examined using OCTA. Results A total of 27.8% and 8.1% of patients had a CRA in one and both eyes, respectively. A total of 174 eyes from 87 patients with a unilateral CRA were examined. Eyes with a CRA showed lower AMD stage (4-step AREDS category; P = 0.037) and a lower prevalence of CNV (23.0% vs. 41.4%; P = 0.024) than those without a CRA. The prevalence of GA and macular vessel density, perfusion density, and foveal avascular zone measured by OCTA were similar in both groups. Conclusions In the eyes with a CRA, AMD stage and prevalence of CNV were lower than those in the eyes without a CRA. However, the effect of the CRA on the macular vascular system remains unclear.

We investigated the associations between retinal vascular geometric measurements and idiopathic epiretinal membrane (ERM). Whether changes in retinal vascular geometry are independent of systemic cardiovascular risk factors was also evaluated. This retrospective, cross sectional study included 98 patients with idiopathic ERM, and 99 healthy age-matched controls. Quantitative retinal vascular parameters were measured from digital retinal fundus photographs using a semi-automated computer-assisted program. Multivariate logistic regression analyses were performed to evaluate associations between retinal vascular geometric parameters and the presence of idiopathic ERM after adjusting for systemic cardiovascular risk factors. There was no significant difference in the baseline characteristics of the two groups, except that the ERM group had a higher proportion of females than the control group. In multivariate regression analyses, after adjusting for age, sex, hypertension, diabetes, hypertriglyceridemia, hypercholesterolemia, and body mass index, wider retinal venular caliber (odds ratio [OR] = 1.058, P P = 0.001) were associated with idiopathic ERM. Idiopathic ERM was associated with alterations in global retinal microvascular geometric parameters, wider retinal venules, and less complex vascular branching patterns, independent of cardiovascular risk factors. These changes may reflect hemodynamic disturbances in the microcirculation of eyes with idiopathic ERM.

Background: To evaluate visual outcomes and changes in fluid after administering monthly anti-vascular endothelial growth factor (anti-VEGF) injections to treat neovascular age-related macular degeneration (nAMD) with subretinal fluid (SRF) and pigment epithelial detachment (PED).Methods: This prospective study included eyes with nAMD previously treated with as needed anti-VEGF injections. The patients were treated with six monthly intravitreal injections of ranibizumab. Quantitative volumetric segmentation analyses of the SRF and PED were performed. The main outcome measures included best-corrected visual acuity (BCVA), and SRF and PED volumes.Results: Twenty eyes of 20 patients were included in this study. At the 6-month follow-up, BCVA and PED volume did not change significantly (P = 0.110 and 0.999, respectively) but the mean SRF volume decreased from 0.53 ± 0.82 mm3 at baseline to 0.08 ± 0.23 mm3 (P = 0.002). The absorption rate of the SRF volume was negatively correlated with the duration of previous anti-VEGF treatment (P = 0.029). Seven of the 20 eyes (35%) showed a fluid-free macula and significant improvement in BCVA (P = 0.036) by month 6.Conclusions: Quantifying the SRF can precisely determinine the patient’s responsiveness to anti-VEGF treatment.

ImportanceUntil now, other than complex neurologic tests, there have been no readily accessible and reliable indicators of neurologic dysfunction among patients with Parkinson disease (PD). This study was conducted to determine the role of fundus photography as a noninvasive and readily available tool for assessing neurologic dysfunction among patients with PD using deep learning methods.ObjectiveTo develop an algorithm that can predict Hoehn and Yahr (H-Y) scale and Unified Parkinson’s Disease Rating Scale part III (UPDRS-III) score using fundus photography among patients with PD.Design, Settings, and ParticipantsThis was a prospective decision analytical model conducted at a single tertiary-care hospital. The fundus photographs of participants with PD and participants with non-PD atypical motor abnormalities who visited the neurology department of Kangbuk Samsung Hospital from October 7, 2020, to April 30, 2021, were analyzed in this study. A convolutional neural network was developed to predict both the H-Y scale and UPDRS-III score based on fundus photography findings and participants’ demographic characteristics.Main Outcomes and MeasuresThe area under the receiver operating characteristic curve (AUROC) was calculated for sensitivity and specificity analyses for both the internal and external validation data sets.ResultsA total of 615 participants were included in the study: 266 had PD (43.3%; mean [SD] age, 70.8 [8.3] years; 134 male individuals [50.4%]), and 349 had non-PD atypical motor abnormalities (56.7%; mean [SD] age, 70.7 [7.9] years; 236 female individuals [67.6%]). For the internal validation data set, the sensitivity was 83.23% (95% CI, 82.07%-84.38%) and 82.61% (95% CI, 81.38%-83.83%) for the H-Y scale and UPDRS-III score, respectively. The specificity was 66.81% (95% CI, 64.97%-68.65%) and 65.75% (95% CI, 62.56%-68.94%) for the H-Y scale and UPDRS-III score, respectively. For the external validation data set, the sensitivity and specificity were 70.73% (95% CI, 66.30%-75.16%) and 66.66% (95% CI, 50.76%-82.25%), respectively. Lastly, the calculated AUROC and accuracy were 0.67 (95% CI, 0.55-0.79) and 70.45% (95% CI, 66.85%-74.04%), respectively.Conclusions and RelevanceThis decision analytical model reveals amalgamative insights into the neurologic dysfunction among PD patients by providing information on how to apply a deep learning method to evaluate the association between the retina and brain. Study data may help clarify recent research findings regarding dopamine pathologic cascades between the retina and brain among patients with PD; however, further research is needed to expand the clinical implication of this algorithm.