Explore the vast range of titles available.

PurposeTo detect the leakage points of central serous chorioretinopathy (CSC) automatically from dynamic images of fundus fluorescein angiography (FFA) using a deep learning algorithm (DLA).MethodsThe study included 2104 FFA images from 291 FFA sequences of 291 eyes (137 right eyes and 154 left eyes) from 262 patients. The leakage points were segmented with an attention gated network (AGN). The optic disk (OD) and macula region were segmented simultaneously using a U-net. To reduce the number of false positives based on time sequence, the leakage points were matched according to their positions in relation to the OD and macula.ResultsWith the AGN alone, the number of cases whose detection results perfectly matched the ground truth was only 37 out of 61 cases (60.7%) in the test set. The dice on the lesion level were 0.811. Using an elimination procedure to remove false positives, the number of accurate detection cases increased to 57 (93.4%). The dice on the lesion level also improved to 0.949.ConclusionsUsing DLA, the CSC leakage points in FFA can be identified reproducibly and accurately with a good match to the ground truth. This novel finding may pave the way for potential application of artificial intelligence to guide laser therapy.

Signs of hypertensive retinopathy can be detected in 2%-14% of the nondiabetic population aged 40 years and older and it has been identified as a risk marker for stroke and cardiovascular mortality. Retinal microvascular changes in hypertensive retinopathy, which can be visualized on fundoscopy, are classified as mild, moderate or malignant, using the Mitchell-Wong system. Generalized retinal arteriolar narrowing and arteriovenous nipping suggest chronic hypertensive damage. More transient effects, reflecting current blood pressure, include focal arteriolar narrowing, retinal hemorrhages, microaneurysms and cottonwool spots. Here, Wong et al examine the case of 23-year-old man with hypertensive retinopathy.

Diabetic retinopathy (DR) is a leading cause of vision-loss globally among type 2 diabetes (T2DM) patients. Information on the economic burden of DR in Singapore is limited. We aim to identify the total annual direct medical costs of DR at different stages, and to examine factors influencing the costs. Four hundreds and seventy T2DM patients who attended the Diabetes Centre in a secondary hospital in Singapore in 2011-2014 were included. Digital color fundus photographs were assessed for DR in a masked fashion. Retinopathy severity was further categorized into non-proliferative DR (NPDR), including mild, moderate and severe NPDR, and proliferative DR (PDR). Medical costs were assessed using hospital administrative data. DR was diagnosed in 172 (39.5%) patients, including 51 mild, 62 moderate and 18 severe NPDR, and 41 PDR. The median cost in DR [2012.0 (1111.2-4192.3)] was significantly higher than that in non-DR patients [1158.1 (724.1-1838.9)] (p<0.001). The corresponding costs for mild, moderate, severe NPDR and PDR were [1167.1 (895.4-2012.0)], [2212.0 (1215.5-3825.5)], [2717.5 (1444.0-6310.7)], and [3594.8.1 (1978.4-8427.7)], respectively. After adjustment, the corresponding cost ratios for mild, moderate, severe NPDR, and PDR relative to non-DR were 1.1 (p = 0.827), 1.8 (p = 0.003), 2.0 (p = 0.031) and 2.3 (p<0.001), respectively. The other factors affecting the total cost include smoking (ratio = 1.7, p = 0.019), neuropathy (ratio = 1.9, p = 0.001) and chronic kidney disease (CKD) (ratio = 1.4, p = 0.019). The presence and severity of DR was associated with increased direct medical costs in T2DM. Our results suggest that preventing progression of DR may reduce the economic burden of DR.

We aim to examine the association of plasma leucine-rich-α-2-glycoprotein 1 (LRG1) with diabetic retinopathy (DR) in type 2 diabetes. At baseline visit, plasma LRG1 levels were assessed using ELISA. Central arterial stiffness was estimated by carotid-femoral pulse wave velocity (PWV). At follow-up visit (median = 3.2 years), digital color fundus photographs were assessed for DR. DR severity was categorized into non-proliferative DR (NPDR) and proliferative DR (PDR). DR was diagnosed in 396 (32.8%) of 1206 patients. DR has higher LRG1 than non-DR (19.5 ± 11.3vs.16.9 ± 8.9 μg/ml, p ≪ 0.001). After adjustment, LRG1 was not associated with DR (OR = 1.2, [95% CI, 0.96–1.30], p = 0.16). LRG1 was higher in PDR (n = 107) than NPDR (n = 270) (23.2 ± 15.4vs.18.1 ± 8.9 μg/ml, n = 270, p ≪ 0.001). After adjustment, with 1-SD increase in LRG1, the relative risk of NPDR and PDR was 0.99 ([0.83–1.18], p = 0.91) and 1.42 ([95% CI, 1.14–1.76], p = 0.002) (p-trend = 0.01), respectively. We didn't observe significant improvement in AUC after adding LRG1 into the model. Baseline PWV mediated 12.0% of the association between LRG1 and PDR (p = 0.03). Baseline plasma LRG1 is associated with PDR, suggesting it maybe a promising biomarker for prediction for advanced proliferative stages of DR. The mediation result indicates the potential benefit of ameliorating central arterial stiffness to prevent PDR.

Accurate detection of ocular signs is essential for early diagnosis of eye diseases, but current AI approaches using facial or external ocular images include non-essential information, compromising performance and patient privacy. We conducted a multinational retrospective study of 2360 eyes from 1180 half-face images of thyroid eye disease patients across five racial groups from five hospitals in three countries. We developed a Dense Squeeze-and-Excitation Network (DSE-Net) to segment eyelid, conjunctiva, lacrimal caruncle, and eyeball, minimizing exposure and enhancing privacy. DSE-Net achieved Dice coefficient of 84.7%, 84.8%, 92.7%, and 95.1%, outperforming seven segmentation models. We then built SegmenView, employing LeNet, AlexNet, ResNet50, and VGGNet16 to detect eyelid edema, conjunctival erythema, caruncle or plica edema, and exophthalmos. SegmenView achieved internal Area Under the Curve (AUCs) of 71.09%, 80.81%, 90.07%, and 82.86%; external AUCs ranging 55.58%–84.29% across two test datasets, outperforming half-face and periocular models. We also compared SegmenView with four privacy-preserving methods, showing its superior ability to balance privacy protection with diagnostic accuracy. Additionally, visualizations based on Gradient-weighted Class Activation Mapping (Grad-CAM) further enhanced the model’s interpretability. Our approach demonstrates high accuracy, generalizability, and potential for lightweight, privacy-preserving ocular sign detection.

This study investigated a Chinese family with congenital posterior polar cataracts linked to the βB2‐R188C mutation. βB2‐crystallin, a key structural component of the lens, is crucial for maintaining lens transparency and stability. We examined the effects of the R188C mutation on βB2‐crystallin's structural stability and resistance to environmental stressors using purified proteins and cellular models. The βB2‐R188C mutant showed poor stability and a tendency to aggregate under physiological and pathological conditions. The mutation disrupted the oligomerization equilibrium, causing dissociation of dimers into monomers. Molecular dynamics simulations and spectroscopic experiments revealed abnormal protein folding induced by the R188C mutation, increasing susceptibility to environmental stressors. Aggregation was observed in both prokaryotic and eukaryotic models under normal conditions, with enhanced severity under environmental stressors. Notably, lanosterol treatment or αB‐crystallin partially reversed aggregation. In summary, the R188C mutation promotes abnormal aggregation by destabilizing βB2‐crystallin and disrupting oligomerization equilibrium, potentially leading to cataract formation. Targeting aggregate formation with small molecules like lanosterol or enhancing molecular chaperone activity offers a promising strategy for cataract prevention and treatment. This study investigated the role of the βB2‐crystallin c.562C>T (p.R188C) mutation in congenital cataract. Clinical pedigree analysis, combined with whole‐exome sequencing (WES), identified this pathogenic variant. Bioinformatic analyses predicted destabilization of the mutant protein’s tertiary structure, potentially impairing its function. In vitro experiments demonstrated that the mutant protein formed aggregates under cellular stress, which were effectively mitigated by treatment with lanosterol and αB‐crystallin. By integrating clinical observations, computational modeling, and experimental validation, this work elucidates how the mutation drives cataract pathogenesis and provides mechanistic insights for pharmacological intervention. These findings suggest a novel strategy for developing non‐surgical therapies targeting protein aggregation in congenital cataracts.

Dear Editor, Khoo Teck Puat Hospital held its first LOTTE (Learning Oriented Teaching in Transdisciplinary Education) elective program for Yong Loo Lin School of Medicine students in April 2021. Faculty critiqued and helped students to refine learning objectives and identify pertinent clinical aspects for elaboration and learning. At an interactive self and peer learning level, students used tools such as the Kahoot! game-based app for peer teaching to optimise learning and increase participation.

Vertical diplopia after blow out fracture repair is not uncommon; we report an unusual case of inferior rectus muscle (IR) transection presenting as a persistent infra-duction deficit after uncomplicated blow out fracture repair. We used multi-positional MRI to diagnose a transected IR with a contracted and posteriorly displaced muscle belly. Infra-duction improved after surgical repair of the transected IR. Multipositional MRI is a novel technology that can be used to assist in the decisive management of persistent post-operative infra-duction deficits and avoid prolonged periods of observation.

To quantify the fundus color on digital color fundus photographs as a surrogate of myopic chorioretinal degeneration in myopic subjects using a novel computer software program. In this cross-sectional study, the following details were recorded in 152 eyes of 78 myopic subjects: Snellen visual acuity, refractive status, ocular biometric parameters, intraocular pressure, color fundus photography, and myopic chorioretinal degeneration index (MCDI) using an automated computer program. The mean age of the subjects was 32.58 ± 11.12 years (range, 21–62 years). Subjects with aged ≥40 years had a significantly higher MCDI when compared with subjects with aged <40 years (MCDI: age ≥ 40 = 0.519; age < 40 = 0.462, P  = 0.000). The MCDI demonstrated a statistically significant and positive relationship with axial length ( r  = 0.47, P  = 0.000) and lens thickness ( r  = 0.26, P  = 0.023). A significant positive relationship was observed between the MCDI and visual acuity ( r  = 0.30, P  = 0.007). The MCDI represents an objective technique for quantitative assessment of chorioretinal degenerative changes in myopic subjects. Older myopic subjects with a relatively higher MCDI and longer axial length may represent a high-risk population who have an increased likelihood of developing visually consequential complications of myopia.

Delayed diagnosis of rare orbital diseases is attributed to limited clinical awareness. Building on prior evidence of multimodal large language models (MLLMs) for detecting common ocular conditions, this study aims to evaluate whether integrating multimodal clinical data can enhance the diagnostic accuracy of MLLMs for rare orbital diseases. We conducted a multinational, multiracial, retrospective study. Two datasets were analyzed: Dataset 1, containing 6,786 single-eye photographs from China, was used to fine-tune a contrastive language-image pre-training (CLIP) for preliminary classification of healthy eyes, orbital diseases, and non-orbital diseases, and to compare its performance against three traditional models and three next-generation models. Dataset 2, comprising 170 participants from China, Singapore, and Thailand, was used to evaluate a MLLM (GPT-4o-Latest). Sequential sensitivity analysis assessed the impact of adding external eye photographs, chief complaints, racial information, and diagnostic reasoning prompts. An AI agent combining the CLIP model with GPT-4o-Latest was further evaluated. The model's ability to generate medical reports and examination recommendations was also assessed. Here we show that the CLIP model achieves 90.21% preliminary detection accuracy, surpassing all baseline models. MLLM detection accuracy improves significantly with the inclusion of multimodal inputs. When relying on external eye images, the top-5 accuracy is 25.68%. The combined agent raises top-5 accuracy to 85.29%. Generated reports and recommendations display high accuracy, readability, completeness, and low potential for harm. Our study demonstrates the potential of MLLM in improving diagnostic accuracy and supporting clinical decision-making for rare orbital diseases.