Interpretable Detection of Diabetic Retinopathy, Retinal Vein Occlusion, Age-Related Macular Degeneration, and Other Fundus Conditions.

Diabetic retinopathy (DR), retinal vein occlusion (RVO), and age-related macular degeneration (AMD) pose significant global health challenges, often resulting in vision impairment and blindness. Automatic detection of these conditions is crucial, particularly in underserved rural areas with limited access to ophthalmic services. Despite remarkable advancements in artificial intelligence, especially convolutional neural networks (CNNs), their complexity can make interpretation difficult. In this study, we curated a dataset consisting of 15,089 color fundus photographs (CFPs) obtained from 8110 patients who underwent fundus fluorescein angiography (FFA) examination. The primary objective was to construct integrated models that merge CNNs with an attention mechanism. These models were designed for a hierarchical multilabel classification task, focusing on the detection of DR, RVO, AMD, and other fundus conditions. Furthermore, our approach extended to the detailed classification of DR, RVO, and AMD according to their respective subclasses. We employed a methodology that entails the translation of diagnostic information obtained from FFA results into CFPs. Our investigation focused on evaluating the models' ability to achieve precise diagnoses solely based on CFPs. Remarkably, our models showcased improvements across diverse fundus conditions, with the ConvNeXt-base + attention model standing out for its exceptional performance. The ConvNeXt-base + attention model achieved remarkable metrics, including an area under the receiver operating characteristic curve (AUC) of 0.943, a referable F1 score of 0.870, and a Cohen's kappa of 0.778 for DR detection. For RVO, it attained an AUC of 0.960, a referable F1 score of 0.854, and a Cohen's kappa of 0.819. Furthermore, in AMD detection, the model achieved an AUC of 0.959, an F1 score of 0.727, and a Cohen's kappa of 0.686. Impressively, the model demonstrated proficiency in subclassifying RVO and AMD, showcasing commendable sensitivity and specificity. Moreover, our models enhanced interpretability by visualizing attention weights on fundus images, aiding in the identification of disease findings. These outcomes underscore the substantial impact of our models in advancing the detection of DR, RVO, and AMD, offering the potential for improved patient outcomes and positively influencing the healthcare landscape.

Types of Descemet Membrane Detachment After Ocular Surface Burns: The Factor Long Been Ignored.

PURPOSE: The aim of this study was to explore types of Descemet membrane detachment (DMD) after ocular surface burns by anterior segment optical coherence tomography. METHODS: This is a pilot, case series, observational study. Patients with DMD after ocular surface burns were enrolled. Ophthalmologic examinations were performed in all patients including slit-lamp photography and anterior segment optical coherence tomography. RESULTS: Three types of DMDs in 9 eyes of 9 patients with ocular surface burns were identified depending on the detachment components involved with the pre-Descemet layer (PDL). Type A was referred as a taut chord that the PDL and Descemet membrane (DM) detached simultaneously but were remained attached to each other, while type B was identified as a wavy line separated from the stroma by a dark slit that demonstrated the detachment of DM from the PDL and stroma. Type C was defined as the DM detached with or without PDL but they were separated from each other. We found that DM and PDL were detached simultaneously in most condition, with type A in 4 cases, type C in 5 cases, and type B in only 1 case. CONCLUSIONS: Our study demonstrated 3 types of DMDs after ocular surface burns and revealed that the limbal involvement and retrocorneal exudations may give clues to DMD in the corresponding areas. DMDs may be neglected for long in patients with extensive limbal involvement in early stages and also play an important role in unstable ocular surface condition until the late stages of conjunctivalization after ocular surface burns.

Vascular flow density in pathological myopia: an optical coherence tomography angiography study.

OBJECTIVES: To investigate vascular flow density in pathological myopia with optical coherence tomography (OCT) angiography. DESIGN: A prospective comparative study was conducted from December 2015 to March 2016. SETTING: Participants were recruited in Beijing Tongren Hospital. PARTICIPANTS: A total of 131 eyes were enrolled, which were divided into three groups: 45 eyes with emmetropia (EM; mean spherical equivalent (MSE) 0.50D to -0.50D), 41 eyes with high myopia (HM; MSE ≤-6.00D, without pathological changes), and 45 eyes with pathological myopia (PM; MSE ≤-6.00D and axial length (AL) ≥26.5 mm, and with pathological changes). MAIN OUTCOME MEASURES: Macular, choriocapillaris and radial peripapillary capillary (RPC) flow densities were measured and compared between groups, and their relationships with AL and best corrected visual acuity (BCVA) were analysed. RESULTS: Significant differences were found in macular, choriocapillaris and RPC flow densities among the three groups (p<0.05). Multiple comparisons revealed that, compared with the EM and HM groups, macular and RPC flow densities of the PM group were significantly decreased (p<0.05), but no significant difference in choriocapillaris flow density was found between the PM and HM groups (p=0.731). Compared with the EM group, retinal flow density in the macular and arcuate fibre region was not decreased in the HM group. In addition, there was a negative correlation between AL and superficial macular flow density (ß=-0.542, p<0.001), deep macular flow density (ß=-0.282, p=0.002) and RPC flow density (ß=-0.522, p<0.001); and a positive correlation between BCVA and superficial macular flow density (ß=0.194, p=0.021), deep macular flow density (ß=0.373, p<0.001), and choriocapillaris flow density (ß=0.291, p=0.001). CONCLUSIONS: Macular and RPC flow densities decreased in pathological myopia compared with high myopia and emmetropia. No significant decrease of retinal flow density in the macular and arcuate fibre region was found in high myopic eyes compared with emmetropic eyes. Moreover, macular and RPC flow densities were negatively related to AL, and macular flow density was positively related to BCVA.

Application of optical coherence tomography angiography in assessment of posterior scleral reinforcement for pathologic myopia.

AIM: To investigate the effect of posterior scleral reinforcement (PSR) on circulation of pathologic myopia eyes with posterior staphyloma by optical coherence tomography angiography (OCTA). METHODS: The study included 30 pathologic myopia eyes with posterior staphyloma which underwent PSR (PSR group) for 6 to 18mo ago, and 30 age and myopia matched eyes without PSR surgery as control group. Macular, choriocapillaris and radial peripapillary capillary (RPC) flow density were measured by OCTA, and the measurements were compared between groups. RESULTS: OCTA found no significant differences in macular flow density between PSR and control groups. For the superficial flow, whole enface flow density (WED), fovea density (FD), and parafoveal density (PD) were 46.55%±5.19% vs 47.29%±4.12% (P=0.542), 31.45%±6.35% vs 31.17%±4.48% (P=0.841), and 48.82%±5.66% vs 49.21%±4.15% (P=0.756) in PSR and control groups, respectively. For the deep flow, WED, FD, and PD were 52.07%±5.78% vs 53.95%±4.62% (P=0.168), 29.62%±6.55% vs 29.50%±6.38% (P=0.940), and 56.93%±6.17% vs 58.15%±5.13% (P=0.407) in PSR and control groups, respectively. The choriocapillary flow density was 61.18±3.25% in PSR group vs 60.88%±2.56% in control group (P=0.692). Also, OCTA found no significant differences in RPCs flow density between PSR and control groups. The optic disc WED, inside disc flow density and peripapillary flow density were 48.47%±4.77% vs 48.11%±4.57% (P=0.813), 45.47%±11.44% vs 46.68%±9.02% (P=0.709), 54.32%±5.29% vs 52.47%±6.62% (P=0.349) in PSR and control groups, respectively. CONCLUSION: OCTA provides a non-invasive and quantitative approach for monitoring macular and papillary blood flow in pathologic myopia. PSR can not improve but may maintain the circulation of pathologic myopia eyes with posterior staphyloma.

Vascular Density in Retina and Choriocapillaris as Measured by Optical Coherence Tomography Angiography.

PURPOSE: To measure the density of the superficial retinal small vessel network (SRSVN), superficial retinal capillary network (SRCN), deep retinal capillary network (DRCN), and choriocapillaris and the size of the foveal avascular zone (FAZ) in the superficial retinal layer in normal eyes. DESIGN: Prospective observational cross-sectional study. METHODS: In healthy Chinese volunteers, the retinal and choroidal vasculature was visualized by split-spectrum amplitude decorrelation angiography-associated optical coherence tomography (RTVueXR Avanti device; Optovue Inc, Fremont, California, USA). RESULTS: Among 105 healthy participants (age: 35.9 ± 13.8 years) mean FAZ measured 0.35 ± 0.12 mm(2), and mean density of SRSVN, SRCN, DRCN, and choriocapillaris was 8.54% ± 0.92%, 31.8% ± 2.6%, 45.8% ± 3.3%, 44.4% ± 3.3%, and 44.5% ± 2.7%, respectively. In multivariate analysis, higher SRSVN density was associated with younger age (P = .001; standardized regression coefficient ß: -0.28), male sex (P = .008; ß: -0.23), lower SRCN density (P < .001; ß: -0.40), and larger mean choriocapillaris vessel diameter (P = .001; ß: 0.30). Higher SRCN density was correlated with male sex (P = .007; ß: -0.19), lower SRSVN density (P < .001; ß: -0.44), and higher density of the radial peripapillary capillary density (P = .004; ß: 0.20). Higher DRCN density was correlated with younger age (P < .001; ß: -0.31), female sex (P = .002; ß: 0.22), higher SRCN density (P < .001; ß: 0.38), and higher choriocapillaris density (P < .001; ß: 0.39). Higher choriocapillaris network density in the central region was associated with higher DRCN density (P < .001; ß: 0.43) and lower radial peripapillary capillary density (P = .005; ß: -0.26). All retinal vascular parameters were not significantly correlated with axial length or subfoveal choroidal thickness. CONCLUSIONS: The density of the macular vascular networks decreases with older age and is independent of axial length and subfoveal choroidal thickness in healthy individuals.