Visual impairments and blindness are major public health concerns globally. Effective eye disease screening aided by artificial intelligence (AI) is a promising countermeasure, although it is challenged by practical constraints such as poor image quality in community screening. The recently developed ophthalmic foundation model RETFound has shown higher accuracy in retinal image recognition tasks. This study developed an RETFound-enhanced deep learning (DL) model for multiple-eye disease screening using real-world images from community screenings. Our results revealed that our DL model improved the sensitivity and specificity by over 15% compared with commercial models. Our model also shows better generalisation ability than AI models developed using traditional processes. Additionally, decision curve analysis underscores the higher net benefit of employing our model in both urban and rural settings in China. These findings indicate that the RETFound-enhanced DL model can achieve a higher net benefit in community-based screening, advocating its adoption in low- and middle-income countries to address global eye health challenges.

Myopia is increasingly prevalent in children. Its association with insufficient sleep has been studied, yielding inconsistent findings. This review aims to assess the association of insufficient sleep with myopia and myopia-related refractive parameters in children. A total of 657 articles were identified, of which 40 were included in the systematic review and 33 were included in the meta-analysis. Results showed that insufficient sleep was significantly associated with an increased prevalence of myopia (odds ratio [OR] = 1.59; 95% confidence interval [CI] = 1.31, 1.95; I² = 99%), and an increased prevalence of high myopia (OR = 3.36; 95% CI = 1.26, 9.00; I² = 96%). Shorter sleep duration was significantly linked to faster changes in axial length (AL) (β = 0.05; 95% CI = 0.02, 0.08; I² = 0%). However, correlation between insufficient sleep and the incidence of myopia, spherical equivalent refraction, corneal curvature radius (CR) and AL/CR were insignificant. Moreover, the effect of insufficient sleep on premyopia and astigmatism was not well-studied. The results of this study suggest that insufficient sleep may be an important risk factor for the development of myopia in school-aged children. Therefore, in addition to ensuring sufficient outdoor activities and reducing near work, it is necessary to inform children and parents about the importance of adequate sleep to mitigate the risk of myopia.

Accurate detection and timely care for patients with high myopia present significant challenges. We developed a deep learning (DL) system enhanced by a self-supervised learning (SSL) approach to improve the automatic diagnosis of myopic maculopathy (MM). Using a dataset of 7,906 images from the Shanghai High Myopia Screening Project and a public validation set of 1,391 images from MMAC2023, our method significantly outperformed conventional techniques. Internally, it achieved 96.8% accuracy, 83.1% sensitivity, and 95.6% specificity, with AUC values of 0.982 and 0.999. Externally, it maintained 89.0% accuracy, 71.7% sensitivity, and 87.8% specificity, with AUC values of 0.978 and 0.973. The model's Cohen's kappa values exceeded 0.8, indicating substantial agreement with retinal experts. Our SSL-enhanced DL approach offers high accuracy and potential to enhance large-scale myopia screenings, demonstrating broader significance in improving early detection and treatment of MM.

Isoegomaketone is a water-soluble natural ketone compound that is commonly present in Rabdosia angustifolia and Perilla frutescens. At present, it is known that isoegomaketone has a wide range of pharmacological activity, but there has been no thorough investigation of its potential targets. As a result, we examined the potential targets of isoegomaketone using the network pharmacology approach. In our study, the TCM Database@Taiwan was utilized to search for the chemical formula. The pharmacological characteristics of isoegomaketone were then evaluated in silico using the Swiss Absorption, Distribution, Metabolism, and Excretion (Swiss ADME) and Deep Learning–Acute Oral Toxicity (DL-AOT) methods, and the potential isoegomaketone target genes were identified using a literature study. Additionally, using the clusterProfiler R package 3.8.1, the Gene Ontology (GO) enrichment analysis and the Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis of target genes were performed. In order to obtain the protein interaction network, we simultaneously submitted the targets to the STRING database. After this, we performed molecular docking with respect to targets and isoegomaketone. Finally, we created visual networks of protein–protein interactions (PPI) and examined these networks. Our results showed that isoegomaketone had good drug-likeness, bioavailability, medicinal chemistry friendliness, and acceptable toxicity. Subsequently, through the literature analysis, 48 target genes were selected. The bioinformatics analysis and network analysis found that these target genes were closely related to the biological processes of isoegomaketone, such as atherosclerotic formation, inflammation, tumor formation, cytotoxicity, bacterial infection, virus infection, and parasite infection. These findings show that isoegomaketone may interact with a wide range of proteins and biochemical processes to form a systematic pharmacological network, which has good value for the creation and use of drugs.

Purpose: To develop deep learning-based networks for the diagnosis of diabetic retinopathy (DR) with cataracts based on infrared fundus images. Setting: Shanghai General Hospital, Shanghai Eye Disease Prevention & Treatment Center, Shanghai, China. Design: Development and evaluation of an artificial intelligence (AI) diagnostic method. Methods: A total of 10 665 infrared fundus images from 4553 patients with diabetes were used to train and test the model. For image quality assessment, left and right eye classification, DR diagnosis and grading, and segmentation of 3 DR lesions, an end-to-end software using EfficientNet and UNet was developed. The accuracy and performance of the software in comparison to human experts was evaluated. Results: The model achieved an accuracy of 75.31% for left and right eye classification, 100% for DR grading and diagnosis tasks, and 73.67% for internal test set, with corresponding areas under the curve (AUCs) of 0.88, 1.00, and 0.89, respectively. For DR lesion segmentation, the AUCs of hemorrhagic, microangioma, and exudative lesions were 0.86, 0.66, and 0.84, respectively. In addition, a contrast test of human-machine film reading confirmed the software's high sensitivity (96.3%) and specificity (90.0%) and consistency with the manual film reading group (κ = 0.869, P < .001). This easily deployable software generated reports quickly and promoted efficient DR screening with cataracts in clinical and community settings. Conclusions: AI-assisted software can perform automatic analysis of infrared fundus images and has substantial application value for the diagnosis of DR patients with cataracts.

Objectives: This study aims to detect characteristic fundus changes in pathological myopia using deep learning (DL)-based analysis of ultra-widefield (UWF) fundus imaging. Methods: Following the exclusion of low-quality images, this cross-sectional study used 1105 UWF images from 543 patients with high myopia to develop the model, along with 293 images from 150 patients with high myopia for external testing. All images were retrospectively collected from patients with high myopia at Shanghai General Hospital and Shanghai Eye Diseases Prevention and Treatment Center between 2018 and 2024. We trained a DL model based on an ophthalmology foundational model to detect myopic maculopathy (MM) and posterior staphyloma (PS). Results: The proposed RETFound-enhanced model demonstrated robust performance. For five-category classification of MM, it achieved 65.4% accuracy and an F1 score of 0.648, outperforming other methods. In three-category MM classification, it achieved 79.4% accuracy and an F1 score of 0.793. For PS detection, the model reached 84.1% accuracy, an F1 score of 0.814 and an area under the receiver operating characteristic curve (AUROC) of 0.886, highlighting its effectiveness as a screening tool. External validation showed consistent performance, with 64.4% accuracy for five-category MM classification, 79.8% accuracy for three-category classification and 81.2% accuracy for PS, confirming robustness across cohorts. Conclusions: This study presents an effective diagnostic model for pathological myopia using UWF fundus imaging and a foundation model. The integration of DL with non-mydriatic UWF fundus imaging demonstrates promising potential for applications in primary healthcare, particularly in underserved areas, enabling accessible screening for high myopia-related fundus changes.