Retinal fluids, detectable through optical coherence tomography (OCT), are key biomarkers for retinal diseases such as diabetic macular edema and age-related macular degeneration, guiding treatment decisions and monitoring response to therapy. Automated segmentation of retinal fluids could support large-scale clinical research and the development of clinical decision support tools. Recent ophthalmic foundation models trained on massive retinal imaging datasets show promise across many downstream tasks, including disease risk prediction and biomarker segmentation, but their performance relative to task-specific models for specialized clinical tasks remains unclear. We compared a task-specific segmentation model (RetiFluidNet) and an ophthalmic foundation model (VisionFM) using a standard benchmarking dataset containing 4,248 OCT images from 48 patients with three retinal diseases. Models were evaluated using three-fold cross-validation and assessed for pixel-level segmentation accuracy and patient-level fluid burden estimation. The task-specific model achieved higher segmentation performance and more consistent fluid quantification across devices. These findings suggest that, for retinal fluid segmentation, specialized task-specific models currently remain more reliable than general-purpose foundation models, highlighting the need for targeted adaptation before clinical deployment.

The increasing prevalence of vision-threatening retinal diseases poses a growing challenge on healthcare systems worldwide, underscoring the need for effective strategies in disease diagnosis and management. Unsupervised artificial intelligence (AI) approaches, such as anomaly detection, offer the potential to identify established and novel retinal imaging biomarkers. They reduce reliance on expert-driven annotations, mitigate bias, and overcome the limitations of predefined disease categories. Moreover, when integrated into clinical workflows, they have the potential to enhance practical applicability by supporting efficient screening, longitudinal disease monitoring, and objective assessment of treatment response.

Several regulatory B lymphocyte (Breg) subsets have been implicated in autoimmune diseases. We investigated Breg subsets in non-infectious uveitis (NIU) and in mouse models of experimental autoimmune uveitis (EAU). Bloods from NIU patients with active disease (n=13), in remission (n=37) and healthy controls (n=10) were immunophenotyped for CD19+CD24hiCD38lo, CD19+CD24hiCD38hi immature (iBreg) and CD19+CD38intCD24int-mature Breg. In NIU, levels of peripheral blood iBreg and serum IL-10 were decreased in active disease vs. remission (P<0.01). In quiescent NIU, iBreg were isolated and, when added at increasing ratios to autologous CD3+T cells, downregulated T cell proliferation. In an EAU model, at peak disease, CD19+CD1dhiCD5+-B10, CD19+CD5+-B1, transitional 2 marginal-zone precursor CD19+CD21hiCD24hiCD23+-T2-MZP and marginal-zone B CD19+CD21hiCD23--MZB Breg subsets were analyzed both phenotypically and functionally. In EAU spleens, all Breg subsets investigated were detectable, but the IL-10-expressing B10 Breg subset was immunosuppressive and inversely correlated with disease severity (P = 0.048, R2 = 0.47). Blood-derived IL-10-expressing B10 and B1 Breg cells were increased in both adjuvant controls and EAU but were significantly reduced in clinically milder EAU. In dissociated retinal cells, only the B10 Breg subset was detectable and increased in EAU. Selective Breg subsets may be immunosuppressive, with iBreg in NIU and IL-10-expressing B10 Bregs in EAU.

Effective treatment of neovascular age-related macular degeneration (AMD) requires targeted inhibition of vascular endothelial growth factor (VEGF) within the retina. However, delivering therapeutic siRNA to the retinal pigment epithelium (RPE), a key source of pathogenic VEGF, remains a major challenge due to ocular barriers and poor cellular tropism. To address this, we developed a retina-targeted delivery system by engineering exosome-liposome hybrid vesicles that encapsulate VEGF-silencing siRNA (Hybrid-siVEGF). The hybrid design leverages the innate homing capability of RPE-derived exosomes for retinal targeting, combined with the high siRNA loading capacity of synthetic liposomes. In vitro, Hybrid-siVEGF showed significantly enhanced uptake by human RPE cells compared to conventional liposomes, leading to the potent and specific knockdown of VEGF expression and the subsequent inhibition of endothelial cell proliferation. In vivo, a single intravitreal injection of Hybrid-siVEGF in a laser-induced choroidal neovascularization mouse model resulted in efficient accumulation within the retina, significant suppression of pathological angiogenesis, preservation of retinal morphology, and restoration of visual function. Our work establishes exosome-liposome hybrids as an effective and targeted platform for ocular siRNA delivery, offering a promising strategy for RNAi-based therapy for AMD and other retinal disorders.

With the rising prevalence of diabetic retinopathy (DR), short videos have become a key source of health information for the public. This study aims to evaluate the quality and interaction performance of DR educational short videos, and explore the impacts of content quality, creator attributes, and platform factors on user interaction. We analyzed 146 DR-themed videos from Douyin and Bilibili, assessed content quality using four international benchmarks (JAMA, mDISCERN, GQS, and PEMAT), and measured interaction via the entropy weight method (EWM). Through non-parametric tests, Spearman correlation analysis, quantile regression, mediation analysis, and moderation analysis, the study found that: overall content quality was moderate, with user interaction indicators showing severe right-skewness; all quality dimensions were significantly positively correlated with each other, while JAMA scores were significantly negatively correlated with the weighted sum of interactions; significant differences existed in multiple indicators across different creator backgrounds, professional titles, and platforms; the impact of quality indicators on interaction was heterogeneous and varied by platform. Conclusions suggest that content creation should be optimized based on platform characteristics, and information seekers should develop habits of rational screening and cross-validation.

Primary cilia are non-motile, microtubule-based cellular protrusions that sense extracellular chemical and mechanical cues, converting them into intracellular signals. Ciliary dysfunction causes ciliopathies including polycystic kidney disease and retinal degeneration. The ciliary membrane is highly enriched in cholesterol, which is essential for the physiological function of ciliary-localized receptors, including G-protein-coupled receptors (GPCRs) and ion channels. Defective de novo cholesterol synthesis, as seen in Smith-Lemli-Opitz syndrome (SLOS), causes typical ciliopathy-like symptoms, suggesting cholesterol deficiency is a core mechanism. We previously identified a pathway where a subset of peroxisomes moves via microtubules to the ciliary base and supplies cholesterol to the ciliary membrane. Furthermore, we found that the mechanosensory cation channel polycystin (PKD1/PKD2) complex, the product of genes underlying Autosomal Dominant Polycystic Kidney Disease (ADPKD), exhibits cholesterol-dependent ciliary localization. Focusing on the ADPKD-associated missense variant p.L517R in the cholesterol-binding site of polycystin-2 (PKD2), we generated mutant mice that displayed ciliopathy spectrums, including situs inversus and polycystic kidneys. The L517R mutant protein showed impaired ciliary localization despite normal channel activity. Intriguingly, exogenous cholesterol administration rescued the ciliary trafficking defect of PKD2 and the expanded epithelial lumen structure in the Zellweger syndrome-modeled cells (peroxisome biogenesis disorder) but not in the PKD2 L517R mutant cells. These findings indicate that the ability of PKD2 to directly bind ciliary membrane cholesterol is essential to prevent ADPKD. This review highlights ciliary cholesterol as a potential therapeutic target for ADPKD.