Disease staging involves the assessment of disease severity or progression and is used for treatment selection. In diabetic retinopathy, disease staging using a wide area is more desirable than that using a limited area. We investigated if deep learning artificial intelligence (AI) could be used to grade diabetic retinopathy and determine treatment and prognosis. The retrospective study analyzed 9,939 posterior pole photographs of 2,740 patients with diabetes. Nonmydriatic 45° field color fundus photographs were taken of four fields in each eye annually at Jichi Medical University between May 2011 and June 2015. A modified fully randomly initialized GoogLeNet deep learning neural network was trained on 95% of the photographs using manual modified Davis grading of three additional adjacent photographs. We graded 4,709 of the 9,939 posterior pole fundus photographs using real prognoses. In addition, 95% of the photographs were learned by the modified GoogLeNet. Main outcome measures were prevalence and bias-adjusted Fleiss’ kappa (PABAK) of AI staging of the remaining 5% of the photographs. The PABAK to modified Davis grading was 0.64 (accuracy, 81%; correct answer in 402 of 496 photographs). The PABAK to real prognosis grading was 0.37 (accuracy, 96%). We propose a novel AI disease-staging system for grading diabetic retinopathy that involves a retinal area not typically visualized on fundoscopy and another AI that directly suggests treatments and determines prognoses.

We developed an unsupervised machine learning algorithm and applied it to big corneal parameters to identify and monitor keratoconus stages. A big dataset of corneal swept source optical coherence tomography (OCT) images of 12,242 eyes acquired from SS-1000 CASIA OCT Imaging Systems in multiple centers across Japan was assembled. A total of 3,156 eyes with valid Ectasia Status Index (ESI) between zero and 100% were selected for the downstream analysis. Four hundred and twenty corneal topography, elevation, and pachymetry parameters (excluding ESI Keratoconus indices) were selected. The algorithm included three major steps. 1) Principal component analysis (PCA) was used to linearly reduce the dimensionality of the input data from 420 to eight significant principal components. 2) Manifold learning was used to further reducing the selected principal components nonlinearly to two eigen-parameters. 3) Finally, a density-based clustering was applied to the eigen-parameters to identify eyes with keratoconus. Visualization of clusters in 2-D space was used to validate the quality of learning subjectively and ESI was used to assess the accuracy of the identified clusters objectively. The proposed method identified four clusters; I: a cluster composed of mostly normal eyes (224 eyes with ESI equal to zero, 23 eyes with ESI between five and 29, and nine eyes with ESI greater than 29), II: a cluster composed of mostly healthy eyes and eyes with forme fruste keratoconus (1772 eyes with ESI equal to zero, 698 eyes with ESI between five and 29, and 117 eyes with ESI greater than 29), III: a cluster composed of mostly eyes with mild keratoconus stage (184 eyes with ESI greater than 29, 74 eyes with ESI between five and 29, and 6 eyes with ESI equal to zero), and IV: a cluster composed of eyes with mostly advanced keratoconus stage (80 eyes had ESI greater than 29 and 1 eye had ESI between five and 29). We found that keratoconus status and severity can be well identified using unsupervised machine learning algorithms along with linear and non-linear corneal data transformation. The proposed method can better identify and visualize the keratoconus stages.

Detection of early clinical keratoconus (KCN) is a challenging task, even for expert clinicians. In this study, we propose a deep learning (DL) model to address this challenge. We first used Xception and InceptionResNetV2 DL architectures to extract features from three different corneal maps collected from 1371 eyes examined in an eye clinic in Egypt. We then fused features using Xception and InceptionResNetV2 to detect subclinical forms of KCN more accurately and robustly. We obtained an area under the receiver operating characteristic curves (AUC) of 0.99 and an accuracy range of 97–100% to distinguish normal eyes from eyes with subclinical and established KCN. We further validated the model based on an independent dataset with 213 eyes examined in Iraq and obtained AUCs of 0.91–0.92 and an accuracy range of 88–92%. The proposed model is a step toward improving the detection of clinical and subclinical forms of KCN.

An oral, multitargeted receptor tyrosine kinase inhibitor, SU11248, inhibits vascular endothelial growth factor receptor 2 (VEG FR2, kinase domain receptor F1K1), platelet-derived growth factor receptor, stem cell factor receptor (Kit), and fetal liver tyrosine kinase 3. In this study, we administered SU11248 beyond the completion of an experimental choroidal neovascularization (CNV) model to determine whether CNV growth can be inhibited in vivo. C57B6J mice were treated with daily oral administration (40, 80, and 150 mg/kg) of SU11248 during 5 days after diode laser endophotocoagulation and subjected to angiographic examination and histological analysis. Intensity of fluorescein leakage from the experimental CNV decreased significantly compared to the control eyes. Histological study also revealed that the volume of CNV membrane was significantly reduced in the SU11248-treated eyes. Oral administration of this compound is potentially beneficial in treating CNV caused by age-related macular degeneration (AMD) and other causal diseases. Considering the tolerance the clinical studies for malignant tumors previously demonstrated, further studies to evaluate the clinical efficacy of the drug for AMD might be indicated.

Purpose To determine changes in multiple cytokine concentrations in the anterior chamber during the induction phase of ranibizumab treatment in patients with neovascular age-related macular degeneration (AMD). Methods This prospective study included 48 treatment-naïve neovascular AMD eyes of 48 patients who received three consecutive monthly injections of ranibizumab at the Japan Community Health Care Organization Tokyo Shinjuku Medical Center between November 2010 and August 2012. We collected ~0.2 mL aqueous humour before the first and third (2 months later) injections. Controls were 80 eyes with cataracts without retinal disease. The cytokines C-X-C motif chemokine ligand 1 (CXCL1), interferon-γ-induced protein 10 (IP-10), C-X-C motif chemokine ligand 12 (CXCL12), C-X-C motif chemokine ligand 13 (CXCL13), monocyte chemoattractant protein 1 (MCP-1), CCL11, C-C motif chemokine ligand 11 (CCL11), interleukin-6 (IL-6), interleukin-10 (IL-10) and matrix metalloproteinase 9 (MMP-9) were analysed using multiplex cytokine assays. Results Mean ages of the patients with AMD and controls were 73 and 75 years, respectively, and 31 (65%) and 37 (46%) subjects were men, respectively. Polypoidal choroidal vasculopathy was found in 27 eyes (56%). Mean concentrations of cytokines in aqueous humour in patients with neovascular AMD before the first and third ranibizumab injections were as follows (in pg/mL): CXCL1, 8.4 and 3.3; IP-10, 110 and 55; CXCL12, 480 and 240; CXCL13, 9.2 and 2.6; MCP-1, 620 and 220; CCL11, 7.1 and 2.8; IL-6, 5.9 and 1.6; IL-10, 0.15 and 0.015 (all p<0.0001), and MMP-9, 0.92 and 1.5 (p=0.0216), respectively. Concentrations of all cytokines decreased significantly after two consecutive ranibizumab injections, except for MMP-9, which increased significantly. Conclusions After two monthly consecutive antivascular endothelial growth factor injections, inflammatory cytokine levels in the aqueous humour of the eyes with AMD were strongly suppressed, while MMP-9 levels increased.

Purpose: Commercially available enzyme-linked immunosorbent assay (ELISA) kits are often used to monitor vascular endothelial growth factor (VEGF) levels in exudative age-related macular degeneration. To test their accuracy, this study performed measurements using the ELISA kits in the presence of anti-VEGF drugs. Methods: The concentrations of bevacizumab, pegaptanib, or ranibizumab at 28 days and aflibercept at 28 and 56 days after an injection were estimated based on previous pharmacokinetic studies. Vascular endothelial growth factor concentrations were measured with two widely used VEGF ELISA kits in the presence of anti-VEGF drugs or control mouse immunoglobulin G (IgG). The monocyte chemotactic protein-1 (MCP-1) ELISA kit was used as a non-VEGF ELISA control kit. Results: The concentrations of aflibercept, bevacizumab, pegaptanib, and ranibizumab were estimated at 0.14 to 7.2, 4.9, 8.6, and 0.11 to 1.1 μg/mL, respectively. ELISA underestimated the VEGF concentration 2- to 100-fold lower in the presence of an anti-VEGF drug, except for pegaptanib, at all VEGF concentrations tested (7.8–1500 pg/mL). Vascular endothelial growth factor at 1000 pg/mL was measured as 92, 150, and 170 pg/mL in the presence of aflibercept (7.2 μg/mL), bevacizumab (4.9 μg/mL), and ranibizumab (1.1 μg/mL), respectively (all P < 0.0001), and the measured VEGF concentration decreased proportionately by 90% to 92% with aflibercept, 85% to 94% with bevacizumab, and 83% to 99% with ranibizumab. The control mouse IgG did not interfere with the measurement of VEGF. Ranibizumab did not affect the measurements with MCP-1 ELISA. Conclusions: Investigators should exercise caution when interpreting measurements of VEGF ELISA in patients being treated with an anti-VEGF drug.