In the United States, gaps in public insurance coverage and high out-of-pocket costs for specialty ophthalmic services contribute substantially to preventable vision loss and disability among underserved populations. We describe the structure, implementation, and early outcomes of ACCESS (Alleviating Costs for Critical Eye Specialty Services), a hospital-based cost-coverage program established at an urban safety-net hospital to address these financial barriers. ACCESS delivers essential vision-restorative treatments frequently excluded from insurance coverage at no cost through hospital support and philanthropic funding. Prospective clinical and demographic data demonstrated marked reductions in visual disability among treated patients and substantial downstream economic benefit relative to program costs. These findings demonstrate that targeted cost-coverage programs embedded within safety-net systems can reduce disparities in vision care, while providing actionable evidence to inform hospital policy development and support broader expansion of insurance coverage for high-impact ophthalmic specialty services.

Purpose: Early screening for eye diseases is critical in low- and middle-income countries where access to care is limited. We investigate whether a confidence-guided, multi-image diabetic retinopathy diagnosis framework can integrate image filtering with confidence-aware predictions for reliable screening at capture. Methods: We develop a multi-image fusion method that aggregates retinal views to improve confidence and balanced accuracy. Our method uses confidence to identify unreliable predictions, prompting retakes when needed. We compare: (1) a cascaded image-quality and disease diagnosis pipeline using a single image per patient, (2) confidence-based prediction, and (3) our confidence-based multi-image fusion pipeline. All methods are evaluated using a RETFoundGreen backbone on the mBRSET (n = 1,234) and BRSET (n = 7,599) datasets. Results: At 70 Conclusions: Human-annotated quality labels are weakly associated with diagnostic performance, and confidence-based filtering consistently outperforms image quality-based cascaded pipelines. Translational Relevance: Using confidence-based multi-image fusion, patients receive more reliable predictions, reducing incorrect diagnoses during screening. The lightweight backbone and single inference pass per image make the framework compatible with low-latency mobile screening systems in resource-limited settings.

The neurosensory retina is one of the most metabolically active tissues in the body and a uniquely accessible extension of the central nervous system, where neuronal and vascular structures can be visualized non-invasively. Its accessibility and highly organized laminar architecture make it a powerful model for studying vascular development and a window into systemic health. Although computational analyses of retinal images have enabled risk assessment for ocular and systemic diseases, most vascular studies rely on two-dimensional frameworks with limited resolution of capillary structure and layer-specific organization. Here, we present a high-resolution three-dimensional (3D) imaging and analysis pipeline enabling quantification of retinal microvasculature and extraction of structural and network metrics across vascular layers. We apply this approach to two mouse models of aberrant retinal vascular development: one with spontaneous postnatal chorioretinal neovascularization and another with disrupted neurovascular lattice formation and layered organization in early life. Across both pathologic contexts, 3D analysis enables detailed characterization of retinal vascular architecture and identifies early vulnerability within the intermediate plexus, the vascular network that bridges the superficial and deep retinal layers, as a sensitive indicator of abnormal remodeling and neovascularization. This framework enables precise characterization of retinal vasculature and establishes a foundation for identifying new retinal biomarkers with potential relevance to neurovascular and systemic disease.

Accurate diagnosis of major eye related diseases such as glaucoma, diabetic retinopathy, diabetic macular edema, pathological myopia, and age related macular degeneration is essential for timely intervention, yet remains challenging in diverse clinical and resource settings. We develop a highly accurate multi-task disease diagnosis model that combines patient metadata with fundus images to diagnose five eye diseases simultaneously. We applied Metafusion, a fusion technique to effectively combine patient metadata (age, gender, history of diabetes and hypertension) with fundus images for diagnosis. In addition, through the development of a novel pre-training method combining supervised and self-supervised losses, we leverage multiple fundus image datasets for diabetic retinopathy and glaucoma to train a multi-task image encoder. The disease diagnosis model demonstrates strong generalization across both lab-captured and smartphone-captured images (Brazilian Multilabel Ophthalmological Dataset, or BRSET, and Mobile BRSET, or mBRSET, datasets respectively). On the BRSET dataset, the disease diagnosis model shows an improvement of 6% in balanced accuracy across all five diseases compared to a model which relies only on images without pre-training enhancements. On the mBRSET dataset there is a 4% improvement in balanced accuracy for diabetic retinopathy and diabetic macular edema. Furthermore, training on a joint (mBRSET+BRSET) dataset preserves performance across both domains, demonstrating model robustness across a range of imaging conditions. This work demonstrates how multimodal data fusion and self-supervised pre-training can improve disease detection accuracy while maintaining high performance across different imaging conditions, which is an important requirement for future ophthalmic diagnostic systems.

Age-related macular degeneration (AMD), glaucoma, diabetic retinopathy (DR), diabetic macular edema, and pathological myopia collectively affect 782 million people worldwide and are among the leading causes of vision loss and blindness. Early screening for these diseases is essential to prevent irreversible vision loss, yet access to medical eye care remains severely limited in low- and middle-income countries (LMICs). To address this gap, we introduce InSight, an AI-powered mobile application that enables simultaneous accurate diagnosis of all five major eye diseases using fundus images captured on a smartphone fundus camera. InSight operates entirely offline without requiring internet connectivity, making it ideal for deployment in resource-limited settings such as LMICs. InSight features a three-stage pipeline consisting of an image quality checker, a disease diagnosis model, and a dedicated DR grading model. By providing real-time feedback during image capture, the app allows users to retake images, which improves the accuracy of diagnosis. A report is also provided to the patient for next steps. The disease diagnosis model is a multi-task model, capable of simultaneously predicting 5 diseases, achieving performance comparable to individual task-specific models while reducing complexity by 5x. The model was trained on a dataset of smartphone-captured fundus images (mobile Brazilian Multilabel Ophthalmological Dataset, or mBRSET), and demonstrated its ability to concurrently detect both diabetic retinopathy and diabetic macular edema. Additionally, training on lab-captured fundus images (BRSET dataset) yielded balanced accuracies of over 80% across all five diseases. Our results highlight the potential of AI in providing accessible early-stage diagnosis in LMICs.

Background/Objectives: Age-related macular degeneration, glaucoma, diabetic retinopathy (DR), diabetic macular edema, and pathological myopia affect hundreds of millions of people worldwide. Early screening for these diseases is essential, yet access to medical care remains limited in low- and middle-income countries as well as in resource-limited settings. We develop InSight, an AI-based app that combines patient metadata with fundus images for accurate diagnosis of five common eye diseases to improve accessibility of screenings. Methods: InSight features a three-stage pipeline: real-time image quality assessment, disease diagnosis model, and a DR grading model to assess severity. Our disease diagnosis model incorporates three key innovations: (a) Multimodal fusion technique (MetaFusion) combining clinical metadata and images; (b) Pretraining method leveraging supervised and self-supervised loss functions; and (c) Multitask model to simultaneously predict 5 diseases. We make use of BRSET (lab-captured images) and mBRSET (smartphone-captured images) datasets, both of which also contain clinical metadata for model training/evaluation. Results: Trained on a dataset of BRSET and mBRSET images, the image quality checker achieves near-100 Conclusions: The InSight pipeline demonstrates robustness across varied image conditions and has high diagnostic accuracy across all five diseases, generalizing to both smartphone and lab captured images. The multitask model contributes to the lightweight nature of the pipeline, making it five times computationally efficient compared to having five individual models corresponding to each disease.

Multiphoton microscopy (MPM) has become a preferred technique for intravital imaging deep in living tissues with subcellular detail, where resolution and working depths are typically optimized utilizing high numerical aperture, water-immersion objectives with long focusing distances. However, this approach requires the maintenance of water between the specimen and the objective lens, which can be challenging or impossible for many intravital preparations with complex tissues and spatial arrangements. We introduce the novel use of cohesive hyaluronan gel (HG) as an immersion medium that can be used in place of water within existing optical setups to enable multiphoton imaging with equivalent quality and far superior stability. We characterize and compare imaging performance, longevity, and feasibility of preparations in various configurations. This combination of HG with MPM is highly accessible and opens the doors to new intravital imaging applications.

The vasculature of the central nervous system is a 3D lattice composed of laminar vascular beds interconnected by penetrating vessels. The mechanisms controlling 3D lattice network formation remain largely unknown. Combining viral labeling, genetic marking, and single-cell profiling in the mouse retina, we discovered a perivascular neuronal subset, annotated as Fam19a4/Nts-positive retinal ganglion cells (Fam19a4/Nts-RGCs), directly contacting the vasculature with perisomatic endfeet. Developmental ablation of Fam19a4/Nts-RGCs led to disoriented growth of penetrating vessels near the ganglion cell layer (GCL), leading to a disorganized 3D vascular lattice. We identified enriched PIEZO2 expression in Fam19a4/Nts-RGCs. Piezo2 loss from all retinal neurons or Fam19a4/Nts-RGCs abolished the direct neurovascular contacts and phenocopied the Fam19a4/Nts-RGC ablation deficits. The defective vascular structure led to reduced capillary perfusion and sensitized the retina to ischemic insults. Furthermore, we uncovered a Piezo2-dependent perivascular granule cell subset for cerebellar vascular patterning, indicating neuronal Piezo2-dependent 3D vascular patterning in the brain.

Retinitis pigmentosa and macular degeneration lead to photoreceptor death and loss of visual perception. Despite recent progress, restorative technologies for photoreceptor degeneration remain largely unavailable. Here, we describe a novel optogenetic visual prosthesis (FlexLED) based on a combination of a thin-film retinal display and optogenetic activation of retinal ganglion cells (RGCs). The FlexLED implant is a 30 µm thin, flexible, wireless µLED display with 8,192 pixels, each with an emission area of 66 µm 2 . The display is affixed to the retinal surface, and the electronics package is mounted under the conjunctiva in the form factor of a conventional glaucoma drainage implant. In a rabbit model of photoreceptor degeneration, optical stimulation of the retina using the FlexLED elicits activity in visual cortex. This technology is readily scalable to hundreds of thousands of pixels, providing a route towards an implantable optogenetic visual prosthesis capable of generating vision by stimulating RGCs at near-cellular resolution.

Purpose To compare the quality of optic nerve photographs from three different handheld fundus cameras and to assess the reproducibility and agreement of vertical cup-to-disk ratio (VCDR) measurements from each camera. Methods Adult patients from a comprehensive ophthalmology clinic and an intravitreous injection clinic in northern Thailand were recruited for this cross-sectional study. Each participant had optic nerve photography performed with each of 3 handheld cameras: the Volk iNview, Volk Pictor Plus, and Peek Retina. Images were graded for VCDR in a masked fashion by two photo-graders and images with > 0.2 discrepancy in VCDR were assessed by a third photo-grader. Results A total of 355 eyes underwent imaging with three different handheld fundus cameras. Optic nerve images were judged ungradable in 130 (37%) eyes imaged with Peek Retina, compared to 36 (10%) and 55 (15%) eyes imaged with the iNview and Pictor Plus, respectively. For 193 eyes with gradable images from all 3 cameras, inter-rater reliability for VCDR measurements was poor or moderate for each of the cameras, with intraclass correlation coefficients ranging from 0.41 to 0.52. A VCDR ≥ 0.6 was found in 6 eyes on iNview images, 9 eyes on Pictor Plus images, and 3 eyes on Peek images, with poor agreement between cameras (e.g., no eyes graded as VCDR ≥ 0.6 on images from both the iNview and Pictor Plus). Conclusions Inter-rater reliability of VCDR grades from 3 handheld cameras was poor. Cameras did not agree on which eyes had large VCDRs.

Importance Telehealth in ophthalmology has traditionally focused on preventive disease screening with limited use in outpatient evaluation. The unique conditions of the COVID-19 pandemic afforded the opportunity to evaluate different implementations of teleophthalmology at scale, providing insight into expanding teleophthalmology care. Objective To compare telehealth use in ophthalmology with other specialties and assess the feasibility of augmenting ophthalmic telehealth encounters with asynchronous testing during the COVID-19 pandemic. Design, Setting, and Participants This quality improvement study evaluated retrospective, longitudinal, observational data from the first 18 months of the COVID-19 pandemic (January 1, 2020, through July 31, 2021) for 881 080 patients receiving care from outpatient primary care, cardiology, neurology, gastroenterology, surgery, neurosurgery, urology, orthopedic surgery, otolaryngology, obstetrics/gynecology, and ophthalmology clinics of the University of California, San Francisco. Asynchronous testing was evaluated for teleophthalmology encounters. Interventions A hybrid care model wherein ophthalmic testing data were acquired asynchronously and used to augment telehealth encounters. Main Outcomes and Measures Telehealth as a percentage of total volume of ambulatory care and use of asynchronous testing for ophthalmic conditions. Results The volume of in-person outpatient visits dropped by 83.3% (39 488 of 47 390) across the evaluated specialties at the onset of shelter-in-place orders for the COVID-19 pandemic, and the initial use of telehealth increased for these specialties before stabilizing over the 18-month study period. In ophthalmology, telehealth use peaked at 488 of 1575 encounters (31.0%) early in the pandemic and returned to mostly in-person visits as COVID-19 restrictions lifted. Elective use of telehealth was highest in gastroenterology, urology, neurology, and neurosurgery and lowest in ophthalmology. Asynchronous testing was combined with 126 teleophthalmology encounters, resulting in change of clinical management for 32 patients (25.4%) and no change for 91 (72.2%). Conclusions and Relevance Telehealth increased across various specialties during the COVID-19 pandemic. Combining teleophthalmic visits with asynchronous testing suggested that this approach is feasible for subspecialty-level evaluation. Additional study is needed to evaluate whether asynchronous testing outside the same institution could provide an effective and lasting approach for expanding the reach of ophthalmic telehealth.

The objective of this study was to compare the sensitivity and specificity of handheld fundus cameras in detecting diabetic retinopathy (DR), diabetic macular edema (DME), and macular degeneration. Participants in the study, conducted at Maharaj Nakorn Hospital in Northern Thailand between September 2018 and May 2019, underwent an ophthalmologist examination as well as mydriatic fundus photography with three handheld fundus cameras (iNview, Peek Retina, Pictor Plus). Photographs were graded and adjudicated by masked ophthalmologists. Outcome measures included the sensitivity and specificity of each fundus camera for detecting DR, DME, and macular degeneration, relative to ophthalmologist examination. Fundus photographs of 355 eyes from 185 participants were captured with each of the three retinal cameras. Of the 355 eyes, 102 had DR, 71 had DME, and 89 had macular degeneration on ophthalmologist examination. The Pictor Plus was the most sensitive camera for each of the diseases (73-77%) and also achieved relatively high specificity (77-91%). The Peek Retina was the most specific (96-99%), although in part due to its low sensitivity (6-18%). The iNview had slightly lower estimates of sensitivity (55-72%) and specificity (86-90%) compared to the Pictor Plus. These findings demonstrated that the handheld cameras achieved high specificity but variable sensitivities in detecting DR, DME, and macular degeneration. The Pictor Plus, iNview, and Peek Retina would have distinct advantages and disadvantages when applied for utilization in tele-ophthalmology retinal screening programs.

Importance Telehealth in ophthalmology has traditionally focused on preventive disease screening with limited use in outpatient evaluation. The unique conditions of the COVID-19 pandemic afforded the opportunity to evaluate different implementations of teleophthalmology at scale, providing insight into expanding teleophthalmology care. Objective To compare telehealth use in ophthalmology with other specialties and assess the feasibility of augmenting ophthalmic telehealth encounters with asynchronous testing during the COVID-19 pandemic. Design, Setting, and Participants This quality improvement study evaluated retrospective, longitudinal, observational data from the first 18 months of the COVID-19 pandemic (January 1, 2020, through July 31, 2021) for 881 080 patients receiving care from outpatient primary care, cardiology, neurology, gastroenterology, surgery, neurosurgery, urology, orthopedic surgery, otolaryngology, obstetrics/gynecology, and ophthalmology clinics of the University of California, San Francisco. Asynchronous testing was evaluated for teleophthalmology encounters. Interventions A hybrid care model wherein ophthalmic testing data were acquired asynchronously and used to augment telehealth encounters. Main Outcomes and Measures Telehealth as a percentage of total volume of ambulatory care and use of asynchronous testing for ophthalmic conditions. Results The volume of in-person outpatient visits dropped by 83.3% (39 488 of 47 390) across the evaluated specialties at the onset of shelter-in-place orders for the COVID-19 pandemic, and the initial use of telehealth increased for these specialties before stabilizing over the 18-month study period. In ophthalmology, telehealth use peaked at 488 of 1575 encounters (31.0%) early in the pandemic and returned to mostly in-person visits as COVID-19 restrictions lifted. Elective use of telehealth was highest in gastroenterology, urology, neurology, and neurosurgery and lowest in ophthalmology. Asynchronous testing was combined with 126 teleophthalmology encounters, resulting in change of clinical management for 32 patients (25.4%) and no change for 91 (72.2%). Conclusions and Relevance Telehealth increased across various specialties during the COVID-19 pandemic. Combining teleophthalmic visits with asynchronous testing suggested that this approach is feasible for subspecialty-level evaluation. Additional study is needed to evaluate whether asynchronous testing outside the same institution could provide an effective and lasting approach for expanding the reach of ophthalmic telehealth.

Telemedicine-based remote digital fundus imaging (RDFI-TM) offers a promising platform for the screening of retinopathy of prematurity. RDFI-TM addresses some of the challenges faced by ophthalmologists in examining this vulnerable population in both low- and high-income countries. In this review, we studied the evidence on the use of RDFI-TM and analyzed the practical framework for RDFI-TM systems. We assessed the novel technological advances that can be deployed within RDFI-TM systems including noncontact imaging systems, smartphone-based imaging tools, and deep learning algorithms.

PURPOSE:The aim of this study is to investigate the efficacy of a mobile platform that combines smartphone-based retinal imaging with automated grading for determining the presence of referral-warranted diabetic retinopathy (RWDR).METHODS:A smartphone-based camera (RetinaScope) was used by non-ophthalmic personnel to image the retina of patients with diabetes. Images were analyzed with the Eyenuk EyeArt® system, which generated referral recommendations based on presence of diabetic retinopathy (DR) and/or markers for clinically significant macular oedema. Images were independently evaluated by two masked readers and categorized as refer/no refer. The accuracies of the graders and automated interpretation were determined by comparing results to gold standard clinical diagnoses.RESULTS:A total of 119 eyes from 69 patients were included. RWDR was present in 88 eyes (73.9%) and in 54 patients (78.3%). At the patient-level, automated interpretation had a sensitivity of 87.0% and specificity of 78.6%; grader 1 had a sensitivity of 96.3% and specificity of 42.9%; grader 2 had a sensitivity of 92.5% and specificity of 50.0%. At the eye-level, automated interpretation had a sensitivity of 77.8% and specificity of 71.5%; grader 1 had a sensitivity of 94.0% and specificity of 52.2%; grader 2 had a sensitivity of 89.5% and specificity of 66.9%.DISCUSSION:Retinal photography with RetinaScope combined with automated interpretation by EyeArt achieved a lower sensitivity but higher specificity than trained expert graders. Feasibility testing was performed using non-ophthalmic personnel in a retina clinic with high disease burden. Additional studies are needed to assess efficacy of screening diabetic patients from general population.