Purpose:To develop and evaluate an unsupervised domain adaptation (UDA) framework for glaucoma classification from fundus images that improves the generalizability of deep learning (DL) models across heterogeneous imaging characteristics and clinical settings. Methods:We developed an adversarial UDA framework that adapts a labeled source domain to an unlabeled target domain by jointly optimizing glaucoma classification and domain discrimination. A total of 6906 fundus images were included: 1422 images (full-view and optic nerve head-cropped) derived from 711 fundus photographs of 520 patients from the University of Illinois Chicago and 5484 images from three public datasets (RIMONE-DL, n = 371; REFUGE, n = 259; LAG, n = 4854). Performance was evaluated across multiple source-target domain pairs. Saliency map analysis compared feature utilization between UDA and standard DL models. Results:Across diverse source-target dataset pairs, the proposed UDA framework improved standard DL classification accuracy by up to 40.0% and area under the curve (AUC) by up to 59.8% in extreme domain shift experiments. UDA also reduced the performance gap to the ideal target-trained model by up to 82.6% for accuracy and 38.1% for AUC. When applied to unseen domains, UDA improved classification accuracy by up to 14.7% relative to standard DL. Conclusions:The proposed UDA framework improves cross-domain generalizability of DL-based glaucoma classification from fundus images. Compared with standard DL models, UDA exhibits feature utilization patterns more aligned with an ideal baseline, reducing overreliance on optic nerve head-specific cues and supporting more robust decision-making under domain shift. Translational Relevance:By enabling glaucoma classification without requiring labeled data from new clinical sites, this UDA framework addresses a key barrier to deploying artificial intelligence-based screening tools across diverse real-world ophthalmic settings.
Purpose:To develop and validate a dual deep learning model (DLM)-based software (BlinkNet-Q) for near real-time quantification of blink rate and incomplete blinks using standard red-green-blue video. Methods:BlinkNet-Q integrated two DLMs: DLM-1 classified open-eye versus complete-blink frames, and DLM-2 differentiated complete from incomplete blinks. This proof-of-concept technical validation study included separate model development and independent testing cohorts. The development cohort comprised 15 participants (30 eyes), yielding 2953 annotated frames for training and internal validation, with participant-level data splitting. The testing cohort included 39 participants (39 eyes) and was used for final evaluation of the BlinkNet-Q. Frame-wise analysis with computer vision-based eye localization enabled automated blink quantification. Performance was assessed using classification metrics, regression-based metrics, intraclass correlation coefficient (ICC), and Bland-Altman analysis. Results:DLM-1 achieved an area under the curve (AUC) of 0.989 (95% confidence interval [CI], 0.932-0.993), and DLM-2 reached an AUC of 0.884 (95% CI, 0.854-0.963). The BlinkNet-Q showed strong agreement with manual counting for blink rate (ICC = 0.973; R² = 0.948; mean absolute error [MAE] = 1.262 blinks/min) and incomplete-blink ratio (ICC = 0.937; R² = 0.876; MAE = 0.059), with small mean biases under standardized recording conditions. Conclusions:BlinkNet-Q demonstrated the technical feasibility of automated blink rate and incomplete-blink quantification under standardized recording conditions, supporting its potential as a software prototype for blink assessment. Further external validation in diverse clinical populations is needed before clinical implementation. Translational Relevance:BlinkNet-Q may support future development of accessible, camera-based blink assessment tools for ocular surface evaluation.
Purpose:To evaluate associations between axial length (AL) and optical coherence tomography (OCT) and OCT angiography (OCTA) parameters, including vessel density (VD) in the superficial and deep macular plexuses, radial peripapillary capillary plexus, and retinal nerve fiber layer (RNFL) thickness. Methods:A cross-sectional observational study included 160 eyes from 160 participants, divided into four AL groups: <24.00, 24.00-24.99, 25.00-25.99, and ≥26.00 mm. The OCT/OCTA parameters were analyzed. Group differences were assessed using Kruskal-Wallis tests with Bonferroni-corrected post hoc pairwise comparisons. Associations between AL and OCT/OCTA parameters were evaluated using correlation and regression analyses. Results:Increasing AL was associated with superior and inferior peripapillary RNFL (pRNFL) thinning, with decreases of -2.24 µm and -3.29 µm per 1-mm AL increase, respectively (P < 0.05). The nasal pRNFL quadrant showed a positive association with AL. RPC VD demonstrated weaker associations, with limited sectoral differences. In the macula, VD decreased as AL increased, with more consistent associations in the deep capillary plexus. Each 1-mm AL increase was associated with 0.29% reduction in the superficial VD and 0.52% reduction in the deep VD. Conclusions:Axial elongation in myopia is associated with region-specific pRNFL changes and reduced macular VD, particularly in the deep capillary plexus. Peripapillary vascular changes were relatively limited compared with structural pRNFL alterations. These findings suggest a distinct pattern of structural and microvascular changes compared with patterns previously reported in glaucoma, although further validation is needed in studies, including glaucoma cohorts. Translational Relevance:Identification of AL-associated OCT/OCTA patterns may improve interpretation of retinal imaging findings in myopic eyes undergoing glaucoma evaluation.
Purpose:To evaluate systemic metabolomic signatures associated with age-related cataract and prioritize signals with age-stable or age-heterogeneous patterns. Methods:We performed two-sample Mendelian randomization (MR) analyses of 249 Nightingale Health metabolomic traits against age-related cataract in FinnGen 12. Higher-confidence MR candidates were defined by inverse-variance weighted (IVW) false discovery rate (FDR) < 0.05, complete directional concordance across seven complementary MR and sensitivity analyses, and nominal support in at least five analyses. MR-prioritized candidates were then triangulated against UK Biobank incident cataract associations. Age-stratum heterogeneity was assessed using tertile-specific UK Biobank estimates and Cochran Q statistics. Results:Among 249 metabolites, 71 met the IVW FDR threshold, and 22 remained after higher-confidence MR filtering. Of these, 13 showed MR-concordant associations with incident cataract in the UK Biobank 500,000 summary layer. Age-stratum assessment prioritized eight metabolites, comprised of five age-stable fatty-acid/lipoprotein-related signals and three age-heterogeneous signals involving glycoprotein acetyls, phenylalanine, and total fatty acids. Conclusions:Using an MR-guided triangulation framework, we prioritized eight systemic metabolomic signals associated with age-related cataract. These signals were comprised of an age-stable fatty-acid/lipoprotein axis and age-heterogeneous inflammatory, amino-acid, and fatty-acid traits, highlighting distinct systemic metabolic patterns that may inform future cataract biomarker and mechanistic studies. Translational Relevance:This study provides a focused set of systemic metabolic candidates for future cataract biomarker validation and prospective risk-stratification research. The integration of MR prioritization, UK Biobank triangulation, and age-stratum heterogeneity assessment offers a transferable framework for studying metabolic signatures in age-related ocular disease.
Purpose:Levodopa (L-DOPA) is an emerging drug repurposing candidate for age-related macular degeneration (AMD). Basic and clinical studies identify a potential role for L-DOPA to prevent and treat neovascular AMD. However, only one prior study has been performed in geographic atrophy (GA). Methods:We performed a retrospective cohort analysis from the Sight Outcomes Research Collaborative database, including eyes with a diagnosis of early- or intermediate-stage AMD in 1 eye. We compared eyes exposed to any dopamine agonist (DA), dopamine receptor D2 agonist, and L-DOPA with eyes with no DA exposure using propensity score matching and survival analysis with multivariable Cox proportional hazard models with new GA development as our primary outcome. Mice were treated with intraperitoneal NaIO3 to induce GA-like pathology. We compared L-DOPA with phosphate-buffered saline-treated controls in both pigmented and albino mice. We quantified retinal thickness by optical coherence tomography, and performed immunofluorescence to quantify choriocapillaris and retinal pigment epithelium density. Results:Any DA, dopamine receptor D2, and L-DOPA exposure were not associated with new-onset GA. NaIO3 treatment caused retinal thinning, choriocapillaris loss, and retinal pigment epithelium degeneration, which were not rescued by L-DOPA treatment in pigmented nor albino mice. Conclusions:In both retrospective database analysis and preclinical models, DA had no effect upon GA pathology. Translational Relevance:Further investigation of levodopa may be more appropriately directed toward neovascular age-related macular degeneration, rather than geographic atrophy, where preclinical, retrospective, epidemiologic, and prospective data suggest a benefit.
Purpose:Sodium fluorescein (SF) is commonly used to observe retinal and choroidal vasculature for clinical or scientific purposes, but few studies focused on concentration of fluorescein, volume injected, and rate of injection in animal experiments. The aim of this study was to investigate the effect of SF on structure and function of the retina in mice. Methods:C57BL/6J mice were injected intraperitoneally with 0.7% saline solution, 5.0% SF, and 10.0% SF, morphological, functional, and molecular changes of the retina were evaluated by fluorescein angiography, electroretinography (ERG), histology, TUNEL assay, immunohistochemistry (IHC), and Western blot. Results:Fluorescein intensity of 10.0% SF-treated mice was higher than 5.0% SF-treated mice. ERG showed early reduction of retinal function, followed progressive restoration of ERG responses. Histology revealed irregular or disorganized features of outer retina after SF administration. TUNEL assay revealed apoptosis in the outer nuclear layer. IHC demonstrated comparable expressions of Rhodopsin, M-opsin, and S-opsin in all groups, but the 10.0% SF-treated mice exhibited the higher immunoreactivity of glial fibrillary acidic protein (GFAP). Molecular analysis showed increased expressions of BAX, NRF2, and HO1, and decreased expression of BCL2 in the retina and retinal pigment epithelial (RPE). Conclusions:Administration of a higher dose of SF could induce some abnormal alterations in the retina, which were not easily distinguishable from pathological changes in the development of ocular diseases in animal models. Translational Relevance:The SF-induced retinal changes that are similar with some histopathological features could affect the authenticity of experimental data, which optimizing the dose of SF should not be ignored in experimental research on mice.
Purpose:To characterize retinal neurovascular alterations in Alzheimer's disease (AD) and mild cognitive impairment (MCI) and examine structure-function relationships. Methods:Eighty-two participants (28 with AD, 21 with MCI, and 33 cognitively normal controls) underwent retinal imaging. Retinal blood flow (RBF) was measured using a function imager. Retinal vessel density (RVD), retinal vessel length density (RVLD), vessel width, and capillary perfusion density (CPD) were quantified with optical coherence tomography (OCT) angiography and retinal tissue volume (RTV) with ultra-high-resolution OCT. Derived metrics included retinal tissue perfusion (RTP), retinal capillary flow index (RCF), and volumetric vessel density. Group comparisons and age-adjusted correlations were performed. Results:Compared with controls, the combined AD + MCI group had significantly lower RBF (3.18 ± 1.02 vs. 4.22 ± 0.79 nL/s; P < 0.001), RCF (0.21 ± 0.07 vs. 0.26 ± 0.04 nL/s/mm; P < 0.001), and RTP (2.89 ± 0.94 vs. 3.86 ± 0.77 nL/s/mm3; P < 0.001). Morphometric indices of the perfused capillary network and retinal tissue volume (RVD, RVLD, CPD, vessel width, RTV) did not differ (all P > 0.05). In AD + MCI, RBF was not associated with structural or morphometric microvascular metrics, whereas controls exhibited age-adjusted correlations between RBF and vessel density and length density. Conclusions:AD and MCI are characterized by reduced retinal perfusion and capillary flow index despite preserved network density and retinal tissue volume, suggesting altered retinal hemodynamic-morphometric relationships. Translational Relevance:Hemodynamic retinal imaging detects early perfusion impairment in AD and MCI before changes in microvascular network density or retinal tissue volume, offering a noninvasive biomarker for detection and therapeutic monitoring.
Purpose:To evaluate tilted crystalline lens morphology in children across different orientations and their correlations with age and spherical equivalent (SE). Methods:This cross-sectional study encompassed 966 eyes of 493 children (5-15 years; SE, -6.00 to +3.00 diopters [D]). The binocular crystalline lens biometrics in 16 orientations were measured using swept-source anterior segment optical coherence tomography. Parameters included the anterior radius of the lens (LAR), posterior radius of the lens (LPR), thickness, and diameter, as well as lens tilt and decentration. Group comparisons were performed among non-, low-, and moderate-myopic eyes. Results:Lens tilt and decentration demonstrated semi-periodic variations (R2 = 0.984 and R2 = 0.789, respectively). Lenses predominantly tilted temporal-inferiorly, with corresponding temporal decentration. The non-myopia group had significantly higher LAR and LPR (more curved) than the low- and moderate-myopia groups (all P < 0.001). As myopia increased, lens tilt consistently decreased across all refractive subgroups (all P < 0.001), whereas other parameters either reached a turning point at approximately -4.00 D or showed no linearity. Age-stratified polynomial analysis revealed that the positive correlation between lens tilt and SE remained strong within the range of -4.18 D to +3.00 D. Conclusions:Although the lens maintains a greater tilt toward the temporal and inferior regions, it exhibits less tilt and becomes more orthotopic in eyes with higher myopia across different age groups in children. Translational Relevance:This study established crystalline lens tilt as a robust, less age-dependent biomarker that could enhance the prediction of myopia progression and facilitate individualized risk assessment in pediatric refractive development.
Purpose:This study aims to investigate the prevalence of refractive error coverage and its multidimensional sociodemographic influencing factors among 1.1 million school-aged students in Hubei Province, China. Methods:This population-based, cross-sectional study (2022-2024) included 1,113,548 students selected via stratified cluster sampling across Hubei Province, China. Standardized clinical assessments comprised distance visual acuity (VA) and noncycloplegic autorefraction. Students were defined as eligible for correction if presenting with functional visual impairment (uncorrected VA >0.2 logMAR) accompanied by myopia (≤-0.50 diopters [D]), hyperopia (≥+2.00 D), or astigmatism (≥0.75 D). Refractive error coverage was calculated as the proportion of eligible students possessing vision-improving glasses. A nested questionnaire substudy (n = 17,014) evaluated multidimensional sociodemographic determinants. Predictors of coverage were identified using both multivariate logistic regression and a random forest machine learning algorithm. Results:Of the 1,113,548 students, 41.26% were eligible for correction, and the overall refractive error coverage was 57.92%. Coverage was significantly positively associated with educational level and inversely with uncorrected visual acuity (UCVA) severity (P < 0.001). Logistic regression revealed higher coverage was significantly associated with female gender, non-left-behind status (odds ratio = 2.30), better academic performance, higher parental education, and parental glasses wear. Random forest analysis identified bilateral UCVA and grade level as the top predictors of coverage. Conclusions:Refractive error coverage among Hubei students remains suboptimal. Improving these rates requires targeted public health interventions prioritizing males, lower-grade students, and rural/left-behind children, alongside initiatives to enhance parental eye health literacy. Translational Relevance:By translating data from over 1.1 million students into predictive models, this study provides a robust, evidence-based foundation to optimize targeted clinical vision screening protocols for vulnerable pediatric populations.
Purpose:The purpose of this study was to evaluate inter- and intra-device agreement of ciliary body (CB) biometric parameters measured by image-registered swept-source anterior segment optical coherence tomography (AS-OCT; ANTERION, Heidelberg Engineering, Heidelberg, Germany) and ultrasound biomicroscopy (UBM; Insight 100, ArcScan). Methods:Patients undergoing anterior segment imaging were prospectively enrolled. Horizontal scans were acquired under standardized illumination. One eye per participant was randomly selected for analysis. Maximum ciliary muscle thickness (CMTMAX), maximum CB thickness (CBTMAX), sulcus-to-sulcus (STS) distance, and scleral spur-to-scleral spur (SSS) distance were manually measured. Eyes with pupillary diameter differences ≥15% were excluded. Inter- and intra-device agreement were assessed using intraclass correlation coefficients (ICCs), linear regression, and Bland-Altman analysis. Results:Seventy-eight eyes from 78 subjects (mean age = 61.9 ± 12.6 years) were analyzed. Inter-device agreement was good to excellent (ICC = 0.60-0.85). Mean absolute inter-device differences were 57.6 ± 36.6 µm for CBTMAX, 40.6 ± 44.7 µm for CMTMAX, 0.28 ± 0.23 mm for SSS, and 0.18 ± 0.16 mm for STS; corresponding mean absolute relative inter-device differences were 4.8% for CBTMAX, 5.9% for CMTMAX, 1.5% for SSS, and 2.5% for STS. Greater baseline CBTMAX was the only factor consistently associated with inter-device differences in CBTMAX or CMTMAX (P < 0.001). Intra-device repeatability was excellent across all parameters for both systems (ICC = 0.78-0.90 for ANTERION and 0.80-0.94 for Insight). Conclusions:Image-registered AS-OCT and UBM provided largely concordant and highly repeatable CB measurements, although systematic inter-device differences across the two modalities suggest they cannot be considered fully interchangeable. Translational Relevance:Modern image-registered swept-source AS-OCT provides a viable, non-contact alternative to UBM for quantitative CB assessment.
Purpose:Inherited retinal diseases are a group of hereditary diseases that cause variable levels of blindness and affect a multitude of adults and children. One such disease is fundus albipunctatus (FA). FA is caused by autosomal recessive retinol dehydrogenase 5 (RDH5) mutations and results in rod dysfunction leading to night blindness and, in a subset of patients, macular degeneration (MD). We previously reported a spontaneous feline model of FA due to an RDH5 missense mutation. The affected cats showed rod dysfunction and a proportion developed degeneration of the area centralis (AC), equivalent to MD in humans. Methods:We used fundus confocal scanning laser ophthalmoscopy and spectral-domain optical coherence tomography imaging, six different electroretinography protocols, immunohistochemistry, and histology/transmission electron microscopy to further characterize this large-animal cat model. Results:In addition to rod dysfunction, cone recovery from intense stimulation was impaired. For RDH5-/- cats that developed AC degeneration, an initial elongation of rod outer segments with disorganization of the distal tips was initially detected in the AC and visual streak, suggestive of impaired shedding/phagocytosis. With progression, photoreceptors degenerated in the AC, matching the MD seen in some human patients. Conclusions:The RDH5-/- cat model recapitulates features of both rod and cone dysfunction seen in human patients with RDH5 mutations. Translational Relevance:The RDH5-/- cat model offers a unique opportunity to further understand the mechanisms of RDH5-associated retinopathies and to investigate potential therapeutic approaches.
Purpose:To assess the within and between-session repeatability of foveal sensitivity measurements using adaptive optics (AO) microperimetry. Methods:Five normal-sighted participants underwent AO microperimetry testing (550-nm circular stimulus) at the preferred retinal locus across two sessions (2 months ± 2 weeks). Two psychophysical paradigms were used: quick estimation of threshold (QUEST) and method of constant stimuli (MOCS). Each paradigm employed both 1.05-arcmin2 and 30.4-arcmin2 stimulus sizes; stimuli were not stabilized in real time. Trial-by-trial responses were fit with a logistic psychometric function to obtain thresholds at a criterion of 78% seen. Thresholds were converted to sensitivity. Statistical comparisons were performed using paired-sample t-tests. Results:Sensitivities measured from QUEST and MOCS procedures were not significantly different (QUEST: 1.05 arcmin2, P = 0.36 for session 1 and P = 0.41 for session 2; MOCS: 30.4 arcmin2, P = 0.14 for session 1 and P = 0.12 for session 2). Data were thus combined across paradigms for tests of repeatability. Sensitivities for data split into two groups within session 1 and separately within session 2 were not significantly different (QUEST: 1.05 arcmin2, P = 0.34 for session 1 and P = 0.51 for session 2; MOCS: 30.4 arcmin2, P = 0.31 for session 1 and P = 0.27 for session 2). Overall, sensitivities across the two sessions were not significantly different (QUEST: 1.05 arcmin2, P = 0.06; MOCS: 30.4 arcmin2, P = 0.81). The coefficients of repeatability across sessions were 1.5 decibels (dB) and 0.4 dB for 1.05 arcmin2 and 30.4 arcmin2, respectively. Conclusions:Sensitivity measurements from AO microperimetry were repeatable within and across sessions. Translational Relevance:Results of the present study will guide future longitudinal studies of disease progression in detecting sensitivity changes using AO microperimetry.
Purpose:To compare the diagnostic performance of two anterior segment optical coherence tomography devices, REVO FC130 and REVO HR, in patients with Fuchs endothelial corneal dystrophy (FECD). Methods:This prospective cross-sectional study included 33 patients with FECD. Each patient was imaged with both devices. Eleven corneal features were graded independently by two evaluators. A third grader resolved disagreements. Agreement between graders and between devices was quantified by calculating the proportion of concordant assessments, including agreement for both detected and non-detected features, with corresponding 95% confidence intervals. Results:REVO HR provided better visualization of key corneal structures, including Descemet's membrane, Bowman's layer, and hyperreflective basal changes. Hyperreflective basal features were detected 11 times more often with REVO HR. Bowman's layer was identified 2.06 times more frequently with REVO HR compared to REVO FC130. Interoperator agreement was high for REVO HR (PropOverall ≥ 0.97 in 10 of 11 features). FC130 showed lower and more variable agreement, especially in low-contrast regions. Interdevice comparisons indicated systematic differences in detection capability. Conclusions:REVO HR showed higher sensitivity and reproducibility in detecting corneal microstructural changes in FECD. Improved image resolution allows better identification of subtle abnormalities, which may support more accurate disease staging and better timing of surgical intervention. For follow-up, the same OCT device should be used consistently. Translational Relevance:This study demonstrates that improved OCT resolution enhances detection of corneal microstructural abnormalities in Fuchs endothelial corneal dystrophy, improving diagnostic precision, disease staging, and clinical decision-making in routine clinical practice.
Purpose:Proliferative vitreoretinopathy (PVR) leads to poor outcomes through recurrent detachment. Early clinical evidence suggests that repeated intravitreal injections of methotrexate (MTX) may offer protective benefits. In this study, we developed a polycaprolactone (PCL)-based drug carrier loaded with MTX and examined its in vitro pharmacodynamics and biological effects in vitro. Methods:PCL was melted and combined with MTX to form a solidified carrier, which was then characterized by electron microscopy. Drug release kinetics were assessed using an ocular pharmacokinetic model. Human retinal pigment epithelium (hRPE) cells and Müller glial cells (MIO-M1) were exposed to the MTX-loaded carrier for 48 hours, and proliferation, metabolism, migration, protein expression, and toxicity were evaluated. Results:Ultrastructural analysis confirmed the typical morphology of the PCL-based matrix. The pharmacokinetic model revealed an initial burst release of MTX followed by a sustained release over approximately 6 months. Exposure to the drug carrier significantly reduced viability and proliferation of both hRPE and MIO-M1 cells, while toxicity was not detected. The drug carrier led to downregulation of fibroblastic protein expression in hRPE, but did not affect MIO-M1. Several other proteins exhibited significant regulation in the proteomics assay. Conclusions:The MTX-loaded PCL carrier maintained a prolonged release profile and demonstrated potential protective effects against PVR in vitro. Further studies are warranted to explore the clinical applicability of this drug delivery system. Translational Relevance:This in vitro study introduces a new PCL-based sustained-release MTX drug carrier that may help address major limitations of current PVR treatments.
Purpose:The purpose of this study was to quantify characteristics of retinal pigment epithelium (RPE) organelle motility in human subjects across the temporal (T) macula. Methods:In 11 healthy volunteers (HVs), we used high-speed adaptive optics-optical coherence tomography (AO-OCT) to image RPE cells from the fovea to 12°T. Organelle motility was calculated in the apical, middle, and basal layers as the time constant (τ) of the exponentially decaying speckle decorrelation function normalized to the photoreceptor layer. We imaged 7 HVs on 3 different days at the fovea and 7.5°T to characterize motility repeatability. We imaged 4 HVs on 2 systems with 830 nm and 1060 nm source wavelengths at 7.5°T to explore motility wavelength dependence. Results:Across the temporal macula, τ decreased from 15.3 (±2.4) to 10.9 (±2.7) seconds (-0.33 seconds/degrees) for the apical RPE, decreased from 9.4 (±1.8) to 8.5 (±1.9) seconds (-0.11 seconds/degrees) for the middle RPE, and was nearly constant (-0.05 seconds/degrees) at 6.7 (±1.0) seconds for the basal RPE at 1060 nm. Motility repeatability was moderate (coefficient of variance [CV] = 10.7% for all subjects and locations). At all RPE layers, τ measured shorter at 830 nm (apical = 6.8 ± 2.2, middle = 4.8 ± 1.0, and basal = 3.8 ± 0.7 seconds) compared with 1060 nm (apical = 10.4 ± 2.8, middle = 6.5 ± 1.6, and basal = 5.0 ± 1.1 seconds, P < 0.005). Conclusions:RPE organelle motility varies across the macula and cell depth. Our normative dataset will aid future studies of RPE-related retinal pathology. Translational Relevance:AO-OCT measured organelle motility is a potential functional biomarker of RPE cell health and ocular disease pathophysiology.
Purpose:To identify, map, and characterize a posterior ocular lymphatic outflow (POLO) pathway from the suprachoroidal space and assess its drainage dynamics and cervical nodal uptake using time-resolved imaging. Methods:Albumin-based tracers (galbumin and CF770/BSA) were delivered into the suprachoroidal space of adult mice under general anesthesia. In vivo imaging included magnetic resonance imaging (MRI) and near-infrared fluorescence imaging using confocal scanning laser ophthalmoscopy (cSLO) of the ocular surface and orbital soft tissues. Ex vivo assessments included in situ head and neck fluorescence imaging 20 minutes after tracer injection and hyperspectral microscopy of ocular and orbital sections. Lymphatic identity was assessed by immunofluorescence staining (podoplanin and vascular endothelial growth factor receptor-3 [VEGFR-3]) in Prox1-tdTomato reporter mice. Results:Following suprachoroidal tracer injection, in vivo MRI and cSLO fluorescence imaging showed preferential drainage toward the nasal orbit. Hyperspectral microscopy showed higher tracer signal in the ipsilateral sclera and orbit than contralateral tissues, and in situ fluorescence imaging detected near-infrared tracer in the ipsilateral cervical lymph node at 20 minutes. Confocal microscopy of the choroid showed co-localization of podoplanin with VEGFR-3 and Prox1 and demonstrated tracer within podoplanin-positive choroidal lymphatic vessels, consistent with a posterior drainage route from the suprachoroidal space to cervical lymph nodes. Conclusions:This study identifies a previously unrecognized POLO pathway and expands ocular drainage physiology beyond the anterior segment to include a posterior lymphatic route through choroidal lymphatic vessels. Translational Relevance:Identification of a POLO pathway reveals a measurable route for fluid clearance, intraocular pressure regulation, and drug delivery, offering new strategies to preserve vision in retinal and posterior segment diseases characterized by fluid accumulation, impaired clearance, or inflammation.
Purpose:The automated classification artificial intelligence (AI) for anterior segment corneal diseases that we developed is capable of classifying images into nine categories, including vision-threatening corneal diseases, such as infectious keratitis; however, it has been trained exclusively on slit-lamp images. We aimed to develop an AI model adaptable to smartphone images by applying transfer learning using smartphone images to the existing AI model. AI trained with transfer learning on smartphone images will be referred to as "Phone-tuned AI." Methods:This study included 2530 images captured using smartphones, collected from multiple collaborating facilities between October 2021 and March 2024. Transfer learning was applied to two existing AI models (You Only Look Once version 5 [YOLOv5] and YOLOX) using these smartphone images, and the accuracy of the Phone-tuned AI was evaluated. Results:The average accuracy for each classification using smartphone images was 93.5% for Phone-tuned AI (YOLOv5) and 67.0% for Original AI (YOLOv5), showing a statistically significant improvement (P = 0.0033). Similarly, the accuracy was 84.2% for Phone-tuned AI (YOLOX) and 78.4% for Original AI (YOLOX), with no significant difference (P = 0.36). When diseases were categorized by urgency, the Phone-tuned AI (YOLOv5) achieved 94.8% for urgent, 89.7% for semi-urgent, 89.4% for routine, and 98.7% for observation-level cases. Conclusions:Phone-tuned AI has the potential to assist in diagnosis and triage in regions with a shortage of ophthalmologists, such as rural areas. Translational Relevance:Transfer learning using smartphone images showed a particularly good fit with YOLOv5, resulting in high diagnostic accuracy and demonstrating strong potential for clinical application.
Purpose:The retina provides a noninvasive window for imaging vascular pulsation in the optic disc. Pulsation characteristics reflect vascular stiffness, hemodynamics, and ocular pressure dynamics, useful in assessing ocular and systemic cardiovascular health. Whereas axial pulsation can be quantified using modified photoplethysmography (PPG), no reliable method exists for quantifying lateral pulsation. Here, we present a method. Methods:Video of the optic disc is acquired over multiple cardiac cycles. Frames are globally aligned and non-rigidly warped to a common space, then averaged to form a template image from which the vessel of interest is segmented. For each frame, the inverse warp is used to compute sub-pixel vessel wall displacements relative to the template. The method and PPG were applied to a vertically oriented vein in one eye from three healthy subjects. Harmonic regression models were used to compare the vessel wall displacement and diameter pulse waves, with the axial pulse wave from PPG. Results:Where the models fitted well (pseudo-R2 ≥ 75%), the lateral pulse waves were either in phase with the axial pulse wave or showed a small phase delay, with the median phase difference ranging from 4.4% to 9.2% of the cardiac cycle (all P ≤ 0.04). The largest difference between left- and right-wall displacement magnitudes (asymmetric displacement) for each subject were 7.6, 25.2, and 6.33 µm respectively. Conclusions:The results demonstrate the efficacy of the method for quantifying lateral pulsation within the optic disc. Translational Relevance:The proposed method demonstrates proof-of-concept feasibility for quantifying lateral pulsation within the optic disc and lays the groundwork for future biomarker investigation.
Purpose:To validate a novel tonometer on ex vivo human eyes. Methods:A device was constructed to apply a known and constant force using a custom tonometer tip, and the resulting applanation image is recorded with a camera. Software automatically measures the applanation area and converts to an intraocular pressure (IOP). Ex vivo human eyes were cannulated, and the IOP manometrically was set to 5 to 45 mm Hg in 5-mm Hg increments. At each IOP, a 5- to 10-second recording was obtained with the device and converted to an IOP. Measured IOPs were compared to the set IOP value. Results:Eight ex vivo human eyes were included in this study. The mean difference (± SD) and mean absolute difference (± SD) between the stable gravity-induced manometric IOP and the measured IOP were -1.2 ± 4.5 mm Hg and 3.1 ± 3.5 mm Hg, respectively. Within the 10- to 30-mm Hg range, the mean difference and absolute difference were 0.7 ± 1.7 mm Hg and 1.4 ± 1.1 mm Hg, respectively. The prototype IOP and the set IOP showed a significant linear correlation across the full range of set IOPs (R2 = 0.90, P < 0.0001), which improved further for the more clinically relevant range of 10- to 30-mm Hg set IOPs (R2 = 0.95, P < 0.0001). Conclusions:Compared to clamped IOP, measurements using a novel handheld automated fixed-force applanation tonometer demonstrated high accuracy, especially within the 10- to 30-mm Hg range. Translational Relevance:We present a novel tonometer validated on ex vivo human eyes, which demonstrates high accuracy and may have applicability to patient care.
