Background: Retinal vascular diseases, including diabetic macular edema (DME), neovascular age-related macular degeneration (nAMD), myopic choroidal neovascularization (mCNV), and branch and central retinal vein occlusion (BRVO/CRVO), are considered vision-threatening eye diseases. However, accurate diagnosis depends on multimodal imaging and the expertise of retinal ophthalmologists. Objective: The aim of this study was to develop a deep learning model to detect treatment-requiring retinal vascular diseases using multimodal imaging. Methods: This retrospective study enrolled participants with multimodal ophthalmic imaging data from 3 hospitals in Taiwan from 2013 to 2019. Eye-related images were used, including those obtained through retinal fundus photography, optical coherence tomography (OCT), and fluorescein angiography with or without indocyanine green angiography (FA/ICGA). A deep learning model was constructed for detecting DME, nAMD, mCNV, BRVO, and CRVO and identifying treatment-requiring diseases. Model performance was evaluated and is presented as the area under the curve (AUC) for each receiver operating characteristic curve. Results: A total of 2992 eyes of 2185 patients were studied, with 239, 1209, 1008, 211, 189, and 136 eyes in the control, DME, nAMD, mCNV, BRVO, and CRVO groups, respectively. Among them, 1898 eyes required treatment. The eyes were divided into training, validation, and testing groups in a 5:1:1 ratio. In total, 5117 retinal fundus photos, 9316 OCT images, and 20,922 FA/ICGA images were used. The AUCs for detecting mCNV, DME, nAMD, BRVO, and CRVO were 0.996, 0.995, 0.990, 0.959, and 0.988, respectively. The AUC for detecting treatment-requiring diseases was 0.969. From the heat maps, we observed that the model could identify retinal vascular diseases. Conclusions: Our study developed a deep learning model to detect retinal diseases using multimodal ophthalmic imaging. Furthermore, the model demonstrated good performance in detecting treatment-requiring retinal diseases.

The aim of this work was to appraise the potential associations of single nucleotide polymorphisms (SNPs) of long non-coding RNA growth arrest-specific 5 (GAS5) with diabetic retinopathy (DR) in a diabetes mellitus (DM) population. Two loci of the GAS5 SNPs (rs55829688 and rs145204276) were genotyped via TaqMan allelic discrimination in 449 non-DR patients and 273 DR subjects. The SNP rs145204276 Del/Del showed a significantly higher distribution in the DR group compared to the non-DR group (AOR: 2.487, 95% CI: 1.424–4.344, p = 0.001). During subgroup analyses, the non-proliferative diabetic retinopathy (NPDR) subgroup demonstrated a significantly higher ratio of the SNP rs145204276 Del/Del (AOR: 2.917, 95% CI: 1.574–5.406, p = 0.001) and Ins/Del + Del/Del (AOR: 1.242, 95% CI: 1.016–1.519, p = 0.034) compared to the non-DR population, while the proliferative diabetic retinopathy (PDR) subgroup did not reveal significant differences in either SNP rs145204276 or rs55829688 distributions compared to the non-DR group. Furthermore, patients with a GAS5 SNP rs145204276 Del/Del showed a significantly shorter DM duration than the wild type (Ins/Ins) (p = 0.021). In conclusion, our findings demonstrate that the GAS5 SNP rs145204276 Del/Del variant is associated with an increased susceptibility to DR in DM patients, particularly in those patients with NPDR.

The glucagon-like peptide 1 (GLP-1) agonist showed anti-hyperglycemic and anti-inflammatory effects, which may retard the risk of external eye disease. The protective effect of GLP-1 agonist and dry eye disease (DED) was found, while the relationship between GLP-1 agonist and other corneal diseases was not clear. Herein, we aim to evaluate the association between the usage of GLP-1 agonists and the development of the following external eye disease in type 2 diabetes mellitus (T2DM) patients. A retrospective cohort study using the National Health Insurance Research Database (NHIRD) of Taiwan was conducted. The T2DM patients were divided into those with GLP-1 treatment and those without GLP-1 treatment and matched with a 1:2 ratio. The main outcomes were the development of dry eye disease (DED), superficial keratitis, and infectious keratitis. The Cox proportional hazard regression was adopted to produce the adjusted hazard ratio (aHR) with a 95% confidence interval (CI) of external eye diseases between groups. There were 115, 54, and 11 episodes of DED, superficial keratitis, and infectious keratitis in the GLP-1 group. Another 280, 168, and 31 events of DED, superficial keratitis, and infectious keratitis were recorded in the control group. The GLP-1 group demonstrated a significantly lower incidence of DED (aHR: 0.853, 95% CI: 0.668–0.989, p = 0.0356) and superficial keratitis (aHR: 0.670, 95% CI: 0.475–0.945, p = 0.0107) compared to the control group. In the subgroup analyses, the correlation of GLP-1 agonist and DED development was more prominent in patients younger than 60 years old (p = 0.0018). In conclusion, the GLP-1 agonist treatments are significantly associated with a lower incidence of subsequent DED and superficial keratitis, while the relationship was not significant between GLP-1 agonist usage and infectious keratitis.

BACKGROUND Retinal vascular diseases, including diabetic macular edema (DME), neovascular age-related macular degeneration (nAMD), myopic choroidal neovascularization (mCNV), and branch and central retinal vein occlusion (BRVO/CRVO), are considered vision-threatening eye diseases. However, accurate diagnosis depends on multimodal imaging and the expertise of retinal ophthalmologists. OBJECTIVE The aim of this study was to develop a deep learning model to detect treatment-requiring retinal vascular diseases using multimodal imaging. METHODS This retrospective study enrolled participants with multimodal ophthalmic imaging data from 3 hospitals in Taiwan from 2013 to 2019. Eye-related images were used, including those obtained through retinal fundus photography, optical coherence tomography (OCT), and fluorescein angiography with or without indocyanine green angiography (FA/ICGA). A deep learning model was constructed for detecting DME, nAMD, mCNV, BRVO, and CRVO and identifying treatment-requiring diseases. Model performance was evaluated and is presented as the area under the curve (AUC) for each receiver operating characteristic curve. RESULTS A total of 2992 eyes of 2185 patients were studied, with 239, 1209, 1008, 211, 189, and 136 eyes in the control, DME, nAMD, mCNV, BRVO, and CRVO groups, respectively. Among them, 1898 eyes required treatment. The eyes were divided into training, validation, and testing groups in a 5:1:1 ratio. In total, 5117 retinal fundus photos, 9316 OCT images, and 20,922 FA/ICGA images were used. The AUCs for detecting mCNV, DME, nAMD, BRVO, and CRVO were 0.996, 0.995, 0.990, 0.959, and 0.988, respectively. The AUC for detecting treatment-requiring diseases was 0.969. From the heat maps, we observed that the model could identify retinal vascular diseases. CONCLUSIONS Our study developed a deep learning model to detect retinal diseases using multimodal ophthalmic imaging. Furthermore, the model demonstrated good performance in detecting treatment-requiring retinal diseases.